WHO Commission on Social Determinants of Health Globalisation and Health: Pathways of Transmission and Evidence of Impact

Globalization and Health Knowledge Network: Research Papers WHO Commission on Social Determinants of Health Globalisation and Health: Pathways of Transmission and Evidence of Impact

Globalization Knowledge Network

Giovanni Andrea Cornia Stefano Rosignoli Luca Tiberti

Pathways of transmission and evidence of impact 1 Preface The Globalization Knowledge Network (GKN) was formed in 2005 with the purpose of examining how contemporary globalisation was influencing social determinants of health. It was one of nine Knowledge Networks providing evidence-informed guidance to the work of the World Health Organization’s Commission on Social Determinants of Health (2005-2008): like most of the Knowledge Networks, its operations were financed by an external funder (in this case, the International Affairs Directorate of Health Canada, Canada’s national ministry of health). The GKN conducted two face-to-face meetings to debate, discuss, outline and review its work, and produced thirteen background papers and a Final Report. These papers and the Final Report underwent extensive internal and external peer review to ensure that their findings and policy inferences accurately reflected available evidence and scholarship.

This GKN publication series was prepared under the general editorship of Ronald Labonté, with assistance from Vivien Runnels and copy-editing provided by Wayne Harding. All views expressed are exclusively those of the authors. A complete list of titles in the publication series appears on the inside back cover of this monograph.

Globalization Knowledge Network Ronald Labonté (Chair) Ted Schrecker (Hub Coordinator)

Layout and Design: Rhonda Carscadden: rhocaDESIGNS.com Globalization and Health Knowledge Network: Research Papers

Globalisation and Health: Pathways of Transmission and Evidence of Impact1,2,3

WHO Commission on Social Determinants of Health

Authors: Giovanni Andrea Cornia University of Florence and Istituto regionale programmazione economica della Toscana (IRPET)

Stefano Rosignoli Istituto regionale programmazione economica della Toscana (IRPET)

Luca Tiberti University of Florence

1 A prior version of this paper was presented on 27 November 2006 at a meeting of the Globalization Knowledge Network of the WHO Commission on Social Determinants of Health held in Kievits Kroon Country Estate, Gauteng, South Africa. The authors would like to acknowledge the comments received on that occasion from Patrick Bond, Corinna Hawkes, Ronald Labonté, David Sanders, Ted Schrecker, and Rolph van der Hoeven. The authors are also much indebted to Roberto De Vogli for his early suggestions on the structuring of the paper, and for comments he provided on its first draft. Many thanks go also to Patrick and Silvyane Guillaumont and other participants to the Development Seminar held at CERDI-University of Clermont Ferrand in December 2006 for the comments provided on the first draft of this paper. Our heartfelt thanks go also to Sheila Marnie who assisted with the editing of the text. 2 The empirical sections of this paper are based on the Globalisation-Health Nexus Database (GHND). A prior version of GHND was compiled in May 2004 with the financial support of the McArthur Foundation. 3 This paper was commissioned and financially supported by the WHO Commission on Social Determinants of Health (http://www.who.int/social_determinants/ en/). Table of contents

Preface 2

Table of acronyms 8

1. Introduction: background, purpose and method of analysis 9

2. Globalisation defined 12 2.1 Endogenous determinants of globalisation 12 2.2 Exogenous (or policy-driven) determinants of globalisation 13 2.3 Actors of exogenous globalisation 14

3. Mortality trends during the recent globalisation 16 3.1 A slow down in the pace of improvement in health status 16 3.2 Changes in the distribution of health and well-being across and within countries 18

4. Mortality models 22 4.1 Material deprivation 23 4.2 Progress in health technology 27 4.3 Acute psychosocial stress 27 4.4 Unhealthy lifestyles 29 4.5 Income inequality, hierarchy and social disintegration 30 4.6. Random shocks 31 4.7 Summary of variables affecting mortality by the five main pathways 32

5. Changes in the social determinants of health: 1980-2005 compared to 1960-80 34 5.1 Household Income 34 5.2 Income subsidies 36 5.3 Economic instability 37 5.4 Income inequality 40 5.5 Trends in formal and informal employment and in unemployment 43 5.6 Inflation and prices of basic goods 44 5.7 Taxation, public expenditure on health care, and approaches to its financing 45 5.8 Migration and family arrangements 48 5.9 Technical progress in health 49

4 Globalization and Health Knowledge Network 5.10 Fertility rates and dependency ratios 51 5.11 Smoking, drinking and obesity 51 5.12 Exogenous developments and random shocks 53

6. Econometric estimates of LEB, IMR, U5MR models 55 6.1 The GHND dataset 55 6.2 Descriptive statistics and bivariate correlation between main variables 57 6.3 Model specification and regression plan 58 6.4 Global results 60 6.5 Regional results 66

7. Simulation of LEB changes due to globalisation & shocks 70

8. Impact of liberalisation-globalisation on the determinants of health 74 8.1 Introduction 74 8.2 Methodological problems and data sources 74 8.3 Results from the literature 76 8.4 Econometric results 77

9. In lieu of conclusions 80

References 83

List of figures

Figure 1. Relation between smoking incidence among 20-55 years old males and LEB 30 Figure 2. Standard deviation of GDP growth rates, 1950-2001 36 Figure 3. Number of financial crises, 1975-2002 38 Figure 4. Changes in male life expectancy at birth in relation to a stress index summarizing changes in unemployment, labour turnover and family completeness, Russia, 1989-1993 40 Figure 5. Trends in the Gini coefficients of income inequality in China and household disposable income in Great Britain, 1960-2003 42 Figure 6. Changes in inequality and suicide rate in transition countries, 1989-97 45

Pathways of transmission and evidence of impact 5 Figure 7. Inflation spline (i.e. the median inflation of countries examined) for the period 1960-2002 45 Figure 8. Trend in the ‘food-price/consumer price’ index by country groups1980-2005 45 Figure 9. Average un-weighted tariff rates (in percent) by region, 1980-1998 46 Figure 10. DPT3 Percentage immunization rate, 1980-1999 49 Figure 11. Trends in alcohol consumption by region 52 Figure 12. Number of conflicts in the last decades 1960-2002 53 Figure 13. Elasticities of the explanatory variables’ coefficients on global estimates for LEB 62 Figure 14. Relation between inequality and LEB, 1960-1980 and 1980-2005 63 Figure 15. Observed and fitted median spline of world LEB 64 Figure 16. LEB elasticities of explanatory variables for four country groups 68 Figure 17. Trend in the observed and estimated of the median spline of the life expectancy at birth in the four sub regions 69 Figure 18. Correlation between the real interest rate and the Gini coefficient of income inequality 77

List of tables

Table 1. Evolution of the Washington Consensus 13 Table 2. Average annual population-weighted a rates of change b of (100-LEB) and IMR, 1960-2004 17 Table 3. Trends in the coefficient of variation and Gini coefficient of the intra-regional and global distribution of 100-LEB, 1960-2000 20 Table 4. Proportional mortality by main disease groups and selected causes, 2001 23 Table 5. Summary of variables affecting health status by main mortality model 33 Table 6. Period GDP/c growth rates* by main regions, 1960-2005 35 Table 7. Share of countries (by region) which experienced negative growth of GDP/c 35 Table 8. Public subsidies /GDP, selected years 37 Table 9. Percentage of households in each income decile receiving different subsidies, Viet Nam 1993 and 1998 37 Table 10. Average standard deviation of GDP/c growth rate by country groups 1960-2005 38

6 Globalization and Health Knowledge Network Table 11. Number of banking and financial crises 39 Table 12. Trends in the Gini coefficients of the distribution of income from the 1950s to the 1990s for 85 developed, developing and transitional economies 41 Table 13. Trends in central government revenue/GDP ratio 43 Table 14. Access to health care and financial difficulties in China, 1993 and 1998 48 Table 15. Cardiovascular interventions* in OECD countries (DDD/1000 inhabitants/day) 49 Table 16. Prevention of mother-to-child transmission with antiretroviral prophylaxis 50 Table 17. Average population-weighted Total Fertility Rates by region, 1960-2005 52 Table 18. Summary of main variables included in GHND and used in regression analysis 56 Table 19. Overlap between geographical and cluster classifications of country groupings 57 Table 20. Descriptive statistics for the variables used in the regression analysis 58 Table 21. Correlation matrix of selected variables included in estimation of the global model 60 Table 22. Results of worldwide regression analysis for 1960-2005 on LEBa, IMR, U5MR 61 Table 23. Results of the regression analysis on LEB, IMR, U5MR for high and middle income countries, 1960-2005 65 Table 24. Results of the regression analysis on LEB, IMR, U5MR for low income and transitional economies, 1960-2005 67 Table 25. Gains and losses of LEB years by 2000 due to policy changes, endogenous changes and random shocks during 1980s-1990s 72 Table 26. Results of the regression of the Lora’s Overall Reform Index on income inequality, GDP/c and volatility of GDP/c 79

Pathways of transmission and evidence of impact 7 Table of Acronyms

AWC Augmented Washington Consensus CEE/CIS Central and Eastern Europe/Commonwealth of Independent States CPI Consumer Price Index CRED Centre for Research on the Epidemiology of Disasters DDD defined daily doses DHS Demographic and Health Survey DPT Diphtheria, Pertussis, Tetanus vaccine EBRD European Bank for Reconstruction and Development ECLAC United Nations Economic Commission on Latin America and the Caribbean EECA Eastern Europe and Central Asia EU European Union FDI Foreign Direct Investments GDP Gross Domestic Product GDP/c Gross Domestic Product per capita GDP-PPP Gross Domestic Product – Public Private Partnerships GHND Globalisation-Health Nexus Database HIV/AIDS Human Immunodeficiency Virus (HIV) and Acquired Immunodeficiency Syndrome (AIDS) IADB Inter-America Development Bank IFI International Financial Institution IMF International Monetary Fund IMF-GFS International Monetary Fund-Government Finance Statistics IMR Infant Mortality Rate IPR Intellectual Property Rights ITC Information Technology and Communication LAC Latin American Countries LEB Life Expectancy at Birth LSMS Living Standards Measurement Study MDGs Millennium Development Goals MENA Middle East and North Africa OECD Organization for Economic Co-operation and Development ORS Oral Rehydration Salts PHC Primary Health Care PHE/GDP pharmaceutical health expenditure/Gross Domestic Product (expenditure as percentage of GDP (PhE/GDP), S&SEA South and South East Asia SOEs State Owned Enterprises SSA sub-Saharan Africa TBC Tuberculosis TFR Total Fertility Rate TNCs Transnational Corporations TRIPS Agreement on Trade-Related Aspects of Intellectual Property Rights U5MR Under-5 Mortality Rate UK United Kingdom UN United Nations UNGASS United Nations General Assembly Special Session USA United States of America USSR Union of Soviet Socialist Republics VAT Value Added Tax WC Washington Consensus WC-AWC Washington Consensus/Augmented Washington Consensus WDI World Development Indicators WHO World Health Organization WTO World Trade Organization

8 Globalization and Health Knowledge Network Globalisation and Health: Pathways of Transmission and Evidence of Impact

1. Introduction: background, purpose and method of analysis

he last quarter century – the years of the re- and steady technical progress in the medical field. The cent globalisation – was one of epochal changes ITC revolution cut information, communication and Twhich impacted favourably on economic growth and transport costs, raised the rate of growth of GDP by health status. The first of these epochal changes was one percent a year in those countries which invested in the end of the Cold War, which is estimated to have it, and facilitated access to health information for mil- freed up a ‘peace dividend’ of 2-4 percent of GDP per lions of people. Finally, the last 10-15 years witnessed year since the early 1990s for the NATO and Warsaw a shift in the development discourse towards a reduc- Pact countries, as well as their allies in the developing tion of infant and child mortality and a reorientation world. Secondly, following the collapse of communism of the international development strategy towards the and military juntas, many countries benefited from a achievement of the Millennium Development Goals ‘democratic dividend’ which led to the spread of par- (MDGs). ticipatory political institutions, democratic values and practices, and a free press. Thirdly, resource use likely The last quarter century also witnessed the emergence became more efficient – ceteris paribus – due to the and spread of an economic paradigm which emphasiz- ‘market dividend’ which came with the introduction es stringent macroeconomic stability, rapid liberalisa- of market reforms in socialist and dirigiste economies. tion of domestic markets, privatization of state-owned Fourthly, with the exception of Africa, most develop- enterprises and public utilities, the removal of barri- ing countries, including Bangladesh, India, China and ers to international trade and financial flows, and the Egypt, enjoyed a considerable ‘demographic dividend’ search for market-based solutions even in the field of due to a sharp deceleration in birth rates and a paral- ‘public goods’. This paradigm aims at creating a global lel rapid growth in the labour force. Fifthly, the last market in which competition among economic agents quarter century benefited from the spread of the ‘in- occurs with only limited government intervention. Its formation and tele-communication (ITC) revolution’ proponents have long maintained that these measures

Pathways of transmission and evidence of impact 9 lead to a reduction in rent-seeking, increased competi- of complete data on the determinants of health and tion, improved export opportunities, and can facilitate the control variables, to the possible effect of omit- the convergence of the incomes and health status of ted variables, and to the limited availability of micro- poor countries towards those of the advanced ones. data. Yet, as noted by Labonté and Schrecker (2006, p11) ‘setting excessively high standards of proof often However, these claims have seldom been validated in provides a good excuse for doing nothing’. Therefore, practice, and economic performance and mortality while fully conscious of the limitations of our analysis, trends have in many cases been disappointing. This we try to arrive at the best evidence-backed conclu- makes the debate on the pace of improvement in sions possible. health status and the reduction of health differentials particularly relevant. Although health status can be – The review of the macro- and micro-scale literature as and indeed has been - affected by non-economic and well as the econometric analyses included in the paper non-policy factors such as the HIV/AIDS pandemic, suggest that – in spite of the ‘dividends’ mentioned natural disasters and conflicts, unsatisfactory trends above - many social determinants of health improved in health status and growing health differentials can at a slower pace in the last quarter century than during also be interpreted as one of the negative effects of the previous two decades, and in some cases showed globalisation policies. On the other hand, sustained negative trends. This does not imply, of course, that all health improvements and health convergence could the blame for these unsatisfactory trends can be placed be interpreted as a broad indication of the success of on globalisation policies. Health changes are also in- liberalisation and globalisation policies (Dollar 2001 fluenced by the structural conditions of each country, and 2004). i.e. its size, international specialisation, distribution of assets, human capital, infrastructure, and quality of This paper attempts to shed some light on the rela- its domestic policies; positive and negative exogenous tion between globalisation policies and health. To this shocks; as well as endogenous changes in the field of end, it discusses changes in health status and its de- democracy, technology, demography and politics. In- terminants over the last quarter century, against the deed, from a theoretical perspective, one could argue background of the trends observed over 1960-80. It that, if introduced under the appropriate conditions, then examines the relationship between globalisation globalisation policies would most likely accelerate policies and the determinants of health. This two stage health improvements. Standard economic theory sug- approach presents considerable methodological and gests that where domestic markets are complete, com- data problems, and the attempt to establish a causal petitive and non-exclusionary, regulatory institutions nexus between globalisation policies and health status strong, asset concentration moderate, access to public cannot be but tentative. The paper uses a variety of health services widespread and social safety nets in analytical tools, including mortality theory; a quanti- place, liberalisation can indeed play a role in reducing tative macro- and – to a lesser extent – micro-analysis rent–seeking, rewarding effort and entrepreneurship, of trends in health status and in its main determinants permitting economies of scale in production, increas- over the last 25 years, in comparison to the preceding ing employment opportunities and improving health twenty years; a review of the many macro- and micro- through improved family earnings and reduced prices prototypical case studies of countries, regions, and for consumer goods. In turn, external liberalisation social groups to illustrate more general points; and a could help to integrate developing nations with good macro-econometric multivariate analysis of the deter- human capital, physical infrastructure and production minants of life expectancy, infant and under-5 mor- potential but funnel domestic markets into the global tality carried out using a database (the Globalisation- market; while policies that facilitate the transfer of Health Nexus Database, GHND) of 136 developed, technology could facilitate a North-South transfer of developing and transition countries for the period drugs, medical equipment and health knowledge. 1960-2005. Despite this multi-pronged approach, the conclusions arrived cannot be defended with the In most cases, however, these conditions are not pres- same degree of confidence as would be obtained in the ent, and the growth, inequality and mortality impact case of controlled experiments. This is due to the lack of premature liberalisation and globalisation can be

10 Globalization and Health Knowledge Network negative, especially if insufficient attention is paid to the influence of existing structural rigidities, incomplete markets and institutions, asymmetric information, persistent protectionism, and the high cost of technology transfer.

Pathways of transmission and evidence of impact 11 Globalisation and Health: Pathways of Transmission and Evidence of Impact

2. Globalisation defined

he term ‘globalisation’ is often used as a catch-all The impact of economic globalisation on health is me- term, which can create considerable confusion, diated through two different sets of factors discussed Tparticularly when discussing the globalisation-health below: firstly through policies which are consciously nexus. In this paper we define globalisation as the aimed at facilitating global economic integration; and process of gradual integration of countries and people secondly through “autonomous” trends which can into the world economy through cross-border flows generate similar or contrary effects, but which are not of goods, services, investments, finance, technology, easy, or in some cases impossible, for policy makers to information, tourists and (to a lesser extent) labour. influence. We refer to the latter as the “endogenous This process makes domestic economic decisions as determinants”, and the former as “exogenous”. This well as local economic and health conditions, increas- distinction is important if we want to identify those ingly conditioned by global conditions, decisions and globalisation policies which lead to improvements in trends. health status and those that worsen it. In the rest of this paper ‘globalisation’ will mainly be understood as Globalisation however has an influence beyond ‘world ‘policy-driven globalisation’. economic integration’: it also impacts on local cul- tures, attitudes, social behaviour and contributes to 2.1 Endogenous determinants of globalisation the formation of global values, as well as to political participation and the spread of democracy. While A first group of factors affecting health include the some of these non-economic effects may have an ef- independent choices of economic agents in the fields fect on health status, they are largely ignored in this of consumption, health innovation, health and repro- paper, which retains a focus on the economic impact ductive behaviour, migration, and so on, i.e. choices of globalisation. that depend on the individual decisions of investors,

12 Globalization and Health Knowledge Network researchers, firms, consumers, voters, health provid- These pro-globalisation measures were preceded and ers, households and communities and that are only made possible by the introduction in the early-mid partially influenced by public policies. While it is im- 1980s of stringent macroeconomic policies which portant to understand their impact on health, it is dif- emphasize low deficits and tax moderation, the lib- ficult, if not impossible, to control them by means of eralisation of domestic markets for goods, labour and policy interventions. finance, the privatisation of state companies and the search for market-based solutions even in the provi- 2.2 Exogenous (or policy-driven) sion of public goods. For instance, the surge of foreign determinants of globalisation direct investments (FDI) in the 1990s often consisted in the acquisition of state-owned enterprises by mul- As noted above, economic globalisation has been driv- tinational firms, and would not have been possible en to a considerable extent by policies adopted since without their prior privatisation and the parallel lib- the early 1980s aimed at liberalising national econ- eralisation of labour markets, which allowed large job omies and integrating them into the global market. cuts to be made in the privatised firms. Likewise, the These policies - mainly applied in formerly inward- surge in short-term portfolio flows of the 1990s would looking, dirigiste developing market economies or in not have occurred without the prior liberalisation of centrally planned economies - are often referred to as domestic financial markets. Thus, both theoretically the ‘Washington Consensus’ (WC) policies. Techni- and empirically, it is impossible to separate out the cally speaking, globalisation entails the removal of bar- effects of globalisation policies from those of domestic riers to international trade in goods and – increasingly liberalisation, and we do not attempt to do so in this – services and government procurement, the opening paper. up to foreign direct investments, the liberalisation of cross-border short-term portfolio and bank-to-bank Because of its unsatisfactory results, the WC package flows, the adoption of a standardised patent regime was enlarged so as to include those ‘improvements regulating technology transfers and intellectual prop- in domestic institutions’ that many argue are neces- erty rights (as embodied in the TRIPS agreement), sary to allow liberalisation and globalisation policies and the simplification of norms on travel, payment to produce their beneficial effects. The new prescrip- systems and all international exchanges, apart from tions require ‘genuine reform ownership’ by the coun- the free circulation of labour. try that introduces them, ‘improved governance’ (i.e.

Table 1: Evolution of the Washington Consensus

Original Washington Consensus ‘Augmented’ Washington Consensus The previous 10 items, plus 1. Fiscal discipline 11. Governance – ownership 2. Reorientation of public expenditure 12. Anti-corruption 3. Tax reform 13. Flexible labour markets 4. Financial liberalisation 14. WTO agreements 5. Unified and competitive exchange rate 15. Financial codes and standards 6. Trade liberalisation 16. “Prudent” capital-account opening 7. Openness to FDI 17. Non-intermediate exchange rate regimes 8. Privatisation 18. Independent central banks/inflation targeting 9. Deregulation 19. Social safety nets 10.Secure Property Rights 20. Targeted poverty reduction

Source: Rodrik (2003)

Pathways of transmission and evidence of impact 13 fighting corruption and red tape), and ‘better- insti cisions. The new welfare economics is based on a more tutions’ such as an independent central bank, secure complex vision of the game between the policy-maker property rights, an independent judiciary and police, and the bureaucracy – seen to be pursuing objectives and ‘safety nets’. This extended approach is generally which differ markedly from those of the citizenry - referred to as the Augmented Washington Consensus and citizens who are driven by ‘rational expectations‘ (Table 1). and who react actively to the introduction of public policies, thus diluting or completely offsetting their Liberalisation-globalisation policies were first imple- impact. mented in the early 1980s in the Anglo-Saxon countries but spread rapidly to the rest of the OECD, and In both these models, all actors in the policy game then to the developing and transitional countries via belong to the country under study. However, in a the policy conditions advocated by the World Bank, globalised economy, the decision-making process is Regional Development Banks and the International strongly influenced by global actors, whose behaviour, Monetary Fund (IMF), and embodied in the count- incentives and motivations also have to be taken into less structural adjustment programs introduced in the account if globalisation is to be harnessed in the inter- 1980s and 1990s. As a result, as noted by Deacon et est of better health outcomes. The main global actors al. (1997), one key facet of globalisation has been the are firstly the international agencies (the IFIs and the globalisation of economic policies. Yet, there have WTO), i.e. components of a world government which been important exceptions to this rule. In all of them, has an increasing influence on national decision-mak- public policy followed a broad market logic but its ap- ing, especially in low- and middle-income countries. plication was selective, and took place after adjusting These agencies have become a major transmission belt policies to local institutions and conditions (Rodrik of the liberal approach to domestic policy making, and 2003). A first group of countries (including Israel, their governance is therefore key for the formulation Turkey, and Uzbekistan) successfully pursued alterna- of alternative globalisation policies. The UN system tive macroeconomic stabilization programmes. A sec- and its agencies are also important actors, especially ond group (including China, Viet Nam and, to some in terms of the normative approaches to development extent, India) broadly adhered to liberal principles and health which they promote. A third major actor are regarding the protection of property rights, market the transnational corporations (TNCs) which domi- competition, macroeconomic stability etc., but did so nate several production and export sectors (including by following a trial-and-error reform path compatible the pharmaceutical sector) through FDIs, ‘outsourc- with local conditions, which ended up differing sub- ing’, and the impact they have on consumer behav- stantially from the WC-AWC blueprint. Other coun- iour, as well as nutritional and health status (Labonté tries (such as Mauritius) liberalised in line with the and Schrecker 2006). Perhaps the most important liberal wish-list, but did so gradually, often maintain- – and least visible – actor of the recent globalisation ing a two-track system with a liberalised sector and process are the ‘money centres’ (banks, pension funds, a protected sector, which allowed them to strengthen hedge funds), rating agencies and global currency their domestic institutions to withstand the shocks markets, which influence the worldwide movement, expected from external liberalisation. Another group maturity and cost of funds - as well as most exchange (including Chile and Malaysia) followed the standard rates – and which have been partly responsible for the blueprint, but adopted different policies in the field growing economic instability which has characterized of capital controls, exchange rate management, public the last quarter century. Between 1992 and 2001, the savings and the role of the state. turnover of foreign exchange markets increased from a daily average of $0.7 trillion to $1.2 trillion, which 2.3 Actors of exogenous globalisation is more than 50 times the daily trade in goods and services and 10 times that in securities (Bhaduri 2005). Traditional welfare economics assume a ‘benign poli- Because of this, governments in countries with an open cy-maker’, who tries to interpret and balance the so- capital account are now compelled to follow policies cial preferences of citizens, firms and social groups, which are in line with the expectations of financial which in turn are seen as passive recipients of her de- markets. Finally, the ‘global media’ is a fourth global

14 Globalization and Health Knowledge Network actor that – de facto – influences policies and institutions through its communication networks. Here too the issue of governance is of utmost importance.

Pathways of transmission and evidence of impact 15 Globalisation and Health: Pathways of Transmission and Evidence of Impact

3. Mortality trends during the recent globalisation era

ur evaluation of the health trends observed one continued to improve in the 1980s and 1990s during the recent globalisation is based on an in both developing and developed countries. How- Oanalysis of the changes observed during this period in ever, the recent literature points to a slowdown in average infant mortality rate (IMR), under-5 mortal- the improvement of this and other health indicators. ity rate (U5MR) and life expectancy at birth (LEB), Wagstaff and Cleason (2004) note that progress in both in absolute terms and in relation to the trends U5MR reduction in the 1990s was slower than in observed in the prior two decades. the 1980s. Deaton (2004) points to a worldwide drop in the rate of decline of child mortality, while 3.1 A slow down in the pace of Deaton and Drèze (2002) as well as other authors improvement in health status have pointed out that in India IMR declined during the 1990s by only 12.5 percent compared to 30 Some authors have claimed that the positive trend percent in the 1980s. A comprehensive evaluation recorded between 1960 and 1980 continued over of health progress by Cornia and Menchini (2006) the last twenty-five years. For instance, Fox (1998) confirmed a widespread slowdown in the rates of im- argues that IMR, LEB and life expectancy at age provement of (100-LEB),4 (a variable which measures

4 Conclusions about changes in health status based on LEB are however biased by the fact that this variable is upper-bounded at around 100 years of age, a fact which implies smaller absolute and relative gains in countries with an already high life expectancy. For this reason, the paper relies on 100-LEB, a variable that measures the life years lost in relation to the maximum attainable. This measure has the advantage of being scale-invariant, i.e. the rate of change of 100-LEB is independent from its value at the beginning of each period. For instance, a 2 year rise in LEB in a country with an initial LEB of 80 years generates a 10 percent fall in 100-LEB that is identical to that generated by a 6 year rise in a country with a LEB of 40. It must be stressed that the selection of an upper bound of 100 years is arbitrary and that replacing it with an upper bound of 90 or 110 years yields different numerical results. However Cornia and Menchini (2006) claim that these would not affect the conclusions regarding the slowdown in the rate of health improvements. While none of the approaches proposed so far for measuring changes in LEB over time (LEB itself, Kakwani’s transformation, logistic transformation, or 100-LEB) are entirely satisfactory, the one cited above is the least unsatisfactory. For more details see Grigoriu (2006).

16 Globalization and Health Knowledge Network the ‘life years lost in relation to the maximum attain- and, to an even greater extent, in the developing coun- able LEB’), over the 1980s and 1990s, and of IMR/ tries in the 1960s, 1970s, and in some cases the 1980s U5MR in the 1990s. following the development of national health systems in newly independent states, and the transfer to the devel- Table 2 below presents evidence confirming these find- oping countries of modern public health technologies. ings. It shows that the rate of decline in (100-LEB) In the socialist countries of Europe such gains were less varied considerably over time and across regions, and pronounced. They, on the contrary, recorded a rise in that the best results were achieved in the 1960s in sub- (100-LEB) in the 1970s due to ‘chronic stress’ (Bobak Saharan Africa (SSA) and Eastern Europe, in the 1970s and Marmot 1996); while the 1990s saw an even more in East Asia, Latin America and the high-income group, pronounced rise due to the sharp increase in cardiovas- in the 1980s in MENA and India, and in the 1990s in cular and violent deaths caused by the ‘acute stress’ as- the South Asian countries other than India, due to the sociated with the highly problematic transition to the rapid IMR fall recorded in Bangladesh. The main mes- market economies (Cornia and Paniccià 2000). Table 2 sage of Table 2 is that there was a steady and generalized also documents the massive decline in life expectancy in decline in the rate of health progress in the 1980s and sub-Saharan Africa caused by the HIV/AIDS pandem- even more so in the1990s,- a decline that is robust to ic, economic stagnation, weakening of health services, the removal of HIV-affected sub-Saharan Africa and of rising inequality and local conflicts. the transition economies from the sample. The second key point is that the slowdown in the Table 2 also shows, however, that there was a slight im- rate of health improvement was most pronounced provement in both variables in 2000-2005. The slow- in the 1990s, possibly pointing to – with the excep- down recorded in the 1980s and 1990s can be juxta- tions mentioned above – the emergence of systemic posed with the rapid gains recorded in the developed problems arising from the policies adopted during this a b Table 2: Average annual population-weighted rates of change of (100-LEB) and IMR, 1960-2004

100-LEB IMR 60-80 80-90 90-00 00-04 60-80 80-90 90-00 00-04 - High income countries -0.84 -0.93 -0.97 -0.35*** -4.8 -3.9*** -3.6 -1.3*** - China -3.21 c -0.63*** -0.45*** -1.01*** -5.5 c -2.5*** -1.4*** -5.8*** - East Asia & Pacific excl. China -1.27 -1.28 -1.03** -1.00 -2.5 -2.1 -3.3*** -2.6* - Europe & Central Asia -0.53 -0.70 -0.32*** -0.70*** -2.1 -2.1 -1.9 -3.3*** - Latin America & Caribbean -1.07 -1.04 -1.02 -0.93 -2.5 -3.5*** -3.6 -2.7*** - Middle East & North Africa -1.17 -1.58*** -1.33*** -1.06*** -2.7 -4.4*** -1.9*** -2.9** - India -0.97 -1.14*** -0.97*** -0.36*** -1.3 -3.4*** -1.6*** -2.4*** - South Asia excl. India -0.94 -1.07 -1.25 -1.26 -1.2 -1.6 -2.1 -1.8 - Sub-Saharan Africa -0.68 -0.22*** 0.64*** -0.05*** -1.7 -0.7 *** -0.7 -0.7

World -1.42 c -0.88*** -0.65*** -0.74* -2.2 c -2.2 -1.0*** -1.6*** World without SSA -1.48 c -0.94*** -0.79*** -0.83 -2.4 c -2.9*** -1.6*** -2.5*** World without SSA and EECA -1.84 c -1.16*** -0.95*** -0.98 -2.6 c -3.2*** -2.1*** -2.8***

Source: authors’ elaboration based on WDI (2006) and UN Population Prospects (2002). Notes: East Asia does not include Japan. a/ The unweighted rates of change in IMR and 100-LEB confirm the trends revealed using the weighted data. b/ The asterisk indicates that the rate of change is different from that of the prior period at the following probability level: *** <0.01, **between 0.01 and 0.05, * between 0.05 and 0.1. The variance of the ‘universe’ used to carry out the test is the population-weighted variance of the decennial rates of change for the countries of each region for the years 1960-2000. c/ These values are influenced by the Chinese famine of 1958-62 and would be smaller if their long term trend value were used. Regional averages are obtained by weighting country data with the data on live births in the case of IMR and the population in that of (100-LEB).

Pathways of transmission and evidence of impact 17 decade, or the influence of other unknown factors. In in the 1990s than in the previous decades. Second, fact, relative to the prior decade, a slowdown is clearly in regions with IMRs below 30-40 per thousand, the evident in China and the East Asian economies, and slowdown may have been due to the elimination of to a lesser extent in Latin America, MENA and India. ‘easy-to-remove’ causes of infant death, and the dif- Interestingly, the high income countries recorded only ficulties encountered in dealing with more complex a modest decline in the pace of reduction of 100-LEB perinatal problems which require more costly solu- in the 1980s and 1990s, suggesting that it is possi- tions. This argument does not, however, do justice to ble to realize steady health gains even at low levels of the comparatively high rates of IMR decline recorded (100-LEB). in high income countries in the 1990s. Third, the stagnation (or rise) of IMR and U5MR in sub-Saharan The IMR trends (as well as the U5MR trends, which Africa was undoubtedly related to the surge in AIDS- are not discussed here for reasons of space) are not very related deaths among infants and young children. Re- different, apart from the fact that for several regions gression analysis carried out by Cornia and Zagonari (Latin America, MENA and India) better results were (2002) shows, for example, that a one percent increase obtained in the 1980s rather that the 1970s. High rates in HIV prevalence among adults leads to a 1.57 and of decline in IMR were recorded in the 1980s for East 0.88 point increase in U5MR and IMR respectively. Asia, Eastern Europe and the more developed coun- In countries which recorded a 20 point rise in HIV tries. In the ‘rest of South Asia’ the greatest improve- prevalence, IMR and U5MR rose respectively by 17 ment was observed in the 1990s, due mainly to the and 32 points. However, these explanations cannot rapid mortality decline recorded in Bangladesh. The fully explain the slowdown in IMR improvement re- reason for the marked improvement in performance in corded in MENA, the economies in transition, China the 1980s was the sharp increase in coverage of public and high-income countries. A broader set of factors health programmes such as child immunisation, oral – possibly including a slower growth of household in- rehydration therapy, pregnancy control, promotion of comes, greater income volatility, and changes in health breast-feeding, and the gradual expansion of primary financing - were likely at play. health care (PHC) systems. 3.2 Changes in the distribution of health and The rapid decline in IMR recorded in Latin America well-being across and within countries and MENA is both remarkable and puzzling, given the recession that hit both regions during the 1980s, The last quarter century has also seen a rise in health and the debt, public finance and inequality crisis expe- inequality within countries, confirming earlier findings rienced by Latin America during the same period. The from the Whitehall Study (Black et al. 1980), as well as gains in child survival achieved in these two regions in other studies which showed that mortality differentials the 1980s were most likely due to the drive to imple- between unskilled and professional workers widened ment PHC measures, steady rises in female education steadily in England and Wales between 1930-32 and and the spread of medical information. In contrast, as 1991-93. Such trends are the exact opposite of those noted by Ahmad et al. (2000), the rate of reduction embodied in national and international targets in this of IMR and U5MR declined sharply in the 1990s in area. For example, the WHO ‘Health for All’ strategy almost all regions and globally. This global decline is set out the following aim in 1984: ‘… by the year observed, even if we remove sub-Saharan Africa and 2000, the actual differences in health status between the economies in transition of Eastern Europe from countries and between groups within countries, should the sample. be reduced by at least 25%’ (Gwatkin 2000).

It has been argued that this deceleration in improve- (i) Evidence of changes in health differentials between ment rates was caused by three factors, first, the de- countries. As suggested by the concavity of the Pres- cline or premature levelling-off of vaccination and oral ton relationship between per capita income and LEB, rehydration rates and coverage of key health services. economic growth should lead in the long term to a As noted by Cutler et al (2005) internationally-sup- convergence in life expectancy between the develop- ported immunisation campaigns were less well funded ing and developed countries. Some analyses do in fact

18 Globalization and Health Knowledge Network support this hypothesis. For instance, Wilson (2001) (not shown for reasons of space) for 1960, 1970, 1980, found that LEB converged across countries starting 1990 and 2000. They found that the trend of both from 1950. Meyer (2001) also suggests that the un- inequality indexes followed a U shape, with the variable weighted cross-country distribution of LEB remained converging until 1990 and diverging thereafter (Table “twin-peaked” over the entire 1960-1997 period, de- 3). However, as also suggested by Goesling and Fire- spite the ‘migration’ of several countries from the left baugh (2004), the divergence in population-weighted to the right peak, and the increase over time in the inequality indexes disappears if sub-Saharan Africa is mode of the two components of this bi-modal distri- removed from the sample, as the fairly rapid conver- bution. Thus a trend towards convergence was evident gence recorded in South Asia (in India and Bangladesh within the low-LEB club of poor countries (at 45-50 in particular) and parts of Central Europe (the Czech years), and also within the high-LEB club of rich ones Republic and Poland) compensates for the divergence (at 75-80 years). Likewise, Micklewright and Stewart registered in most countries of the former Soviet Union (1999) found that the standard deviation of the distri- and nations such as Iraq, North Korea, Haiti etc. Final- bution of U5MR of the 15 members of the European ly it should be noted that, when the coefficients of dis- Union (EU) declined over 1970-95 by 90 percent as persion are computed without weighing for population Southern Europe moved closer to those of Northern size, the global divergence in health inequality persists Europe due, inter alia, to the structural and regional in the 1990s, even when sub-Saharan Africa is removed subsidies (equivalent to 3-4 percent of the GDP of from the sample (see bottom of Table 3 for 100-LEB). Southern nations) provided by the EU Cohesion All this implies that –- the derailment of the long-term Fund, as well as the European Union’s adoption of convergence in life expectancy (in terms of countries compulsory standards in many health–related areas. rather than people) predicted by Preston (1976) and Wilson (2001) was not just due to the spread of HIV/ However, most analyses point to a growing divergence AIDS and the adverse economic and conditions in Af- in health status between countries. For the 1992-2000 rica. This is an important point which has already been period Goesling and Firebaugh (2004) found a po- noted by McMichael et al. (2004) and which merits larization in the cross country distribution of LEB, greater attention by policy-makers. which they attribute to the HIV/AIDS epidemics in sub-Saharan Africa. McMichael et al. (2004) identify Table 3 also shows that while the intra-regional dis- three different trends in LEB. In the high-income na- persion in (100-LEB) followed a U-shaped trend in tions the LEB trend is plateauing, while in the middle- Eastern Europe, it diverged steadily in all other re- income countries it is converging rapidly towards that gions with the exception of East Asia and Western of the advanced nations. In contrast, a third group of Europe, where a clear convergence was registered. at least 42 countries (including sub-Saharan African Interestingly, in SSA there was a ‘downward conver- countries and the economies in transition, but also the gence’ in 100-LEB between 1990 and 2000, as the Bahamas, Dominican Republic, Fiji, Haiti, Hondu- countries which suffered the biggest losses of life ex- ras, Iraq and North Korea) had a lower life expectancy pectancy (South Africa, Zimbabwe, Botswana and in 2001 than in 1960, 1980 or 1990. The authors ar- etc) had previously recorded the largest gains. Ex- gue that the usual explanations of health convergence cept for East Asia, which is dominated by China, (i.e. the rapid fall in deaths due to infectious diseases these trends are more pronounced when the coef- in poorer countries and the slower decline in mortality ficient of variation and the Gini coefficient are -com due to chronic diseases in rich countries) have to be puted without weighing for the population of each broadened to take into account new life-threatening country. challenges. The analysis of health convergence carried out on IMR To analyze the growing differentials in average life du- and U5MR confirms even more clearly than in the ration, Cornia and Menchini (2006) calculated the case of (100-LEB), that the recent global and intra- weighted and unweighted coefficient of variation and regional health gains were distributed in an increas- Gini coefficient of the global and regional distribu- ingly less egalitarian way, particularly throughout the tions of country 100-LEB (Table 3) and IMR/U5MR 1990s. This conclusion is robust to the choice of the

Pathways of transmission and evidence of impact 19 Table 3: Trends in the coefficient of variation and Gini coefficient of the intra-regional and global distribution of 100-LEB, 1960-2000

100- LEB Coefficient of variation (pop. weighted values) Gini coefficient (pop. weighted values) 1960 1970 1980 1990 2000 1960 1970 1980 1990 2000 East Asia and Pacific (22) 0.18 0.16 0.16 0.15 0.15 7.98 7.23 7.48 6.51 6.17 L. America & Caribbean (32) 0.12 0.12 0.12 0.12 0.13 6.54 6.17 6.4 6.2 6.49 Middle East & N.Africa (20) 0.10 0.10 0.10 0.12 0.14 3.61 4.06 4.99 6.06 6.78 Sub Saharan Africa (45) 0.07 0.08 0.09 0.11 0.09 3.7 4.33 5.1 5.9 4.59 South Asia (7) 0.05 0.06 0.07 0.07 0.08 1.61 1.89 2.11 2.12 1.94 Eastern Europe & C.Asia (29) 0.15 0.13 0.08 0.05 0.10 6.81 5.55 4.18 2.88 5.44 Western Europe (18) 0.05 0.04 0.04 0.04 0.04 2.29 2.12 2.29 2.10 1.92 North America (2) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.

World (175) 0.27 0.24 0.24 0.23 0.27 15.2 13.32 13.19 12.86 14.18 World excl. SSA (130) 0.28 0.22 0.22 0.20 0.19 15.63 12.57 11.98 10.87 10.31 World excl.SSA & 0.26 0.22 0.22 0.20 0.20 14.47 12.21 12.32 11.31 10.67 EECA (101) Memo item: unweighted values World (175) 0.27 0.27 0.28 0.30 0.35 15.4 15.6 16.0 16.8 19.4 World excl. SSA (130) 0.25 0.24 0.23 0.23 0.24 14.2 13.3 12.5 11.8 12.5

Source: Cornia and Menchini (2006) on WDI (2004). Notes: the regional aggregates include only developing countries (e.g. East Asia does not include Japan); the number of countries in each area is given in parenthesis. Notes: n.a. : not applicable.

inequality and health indicator, and whether or not we allel increase in health differentials, although lack of apply population weights to the national data. data makes it difficult to test this assumption. Grow- ing U5MR differentials by income level are reported (ii) Evidence of within-country divergence in health status. by Minujin and Delamonica (2003) for the 1980s Improvements in nationwide health indicators may and 1990s on the basis of an analysis of Demographic conceal vastly different rates of improvements within and Health Surveys (DHS) for 24 developing coun- different social groups, whether these are defined in tries. The ratio of the U5MR of children of families terms of gender, rural-urban residence, region, income belonging to the bottom 20 percent relative to that class, ethnic group, education level, labour market sta- of children of households belonging to the top 20 tus of the head of the household, and so on. The rather percent worsened in 11 of the 24 countries included limited evidence of changes over time in within-coun- in the study, remained constant in 10 and improved try mortality differentials over the last quarter century in three. A similar worsening of U5MR differentials is reviewed below. was observed not only in countries where the average U5MR worsened, but also in several of those where There is ample evidence of mortality differentials pro- the average fell, suggesting that most of the reduction vided in the literature (Ruzicka et al. 1989, Wagstaff was concentrated in the top income group. 2000, Cornia and Paniccià 2000) especially in unequal societies. This is a point that needs underscor- Broadly similar results were obtained by Cornia and ing, as income inequality appears to have grown in Menchini (2006), who analyze changes in IMR and many countries during the last twenty-five years (see U5MR differentials between the early 1990s and the below), suggesting that there may have also been a par- early 2000s using DHS data. IMR for different in-

20 Globalization and Health Knowledge Network come groups are calculated using an ‘asset index’ (a proxy for family income), and differentials were found to have increased in 60 percent of the cases analysed. Again improvements in average IMR were accompanied by growing (or unchanged) IMR differentials, suggesting that much of the decline in IMR took place in the middle class and the rich income groups. Similar results were obtained for the rural-urban IMR differential, which broadened in 11 cases, remained unchanged in four, and narrowed in six, suggesting that most of the IMR decline was recorded in urban areas.

Evidence of rising IMR differentials for China is given by Zhang and Kanbur (2003) in their study which shows that the ratio of rural/urban IMR rose from 1.5 to 2.1, between 1981 and 1995, and that the female/ male IMR ratio rose from 0.9 to 1.3, while the regional variation in health outcomes rose sharply. The authors link these trends to the fiscal decentralisation of 1978, the dissolution of the communes, the diffusion of private medical practices since 1984 and the freedom granted to urban-based state-owned enterprises (SOEs) to lay off workers and cut health subsidies. Finally, mortality differentials among adults rose sharply in the transition economies, both in those that experienced a fall in LEB (such as the Russian Federa- tion) and those that recorded an improvement (such as the Czech Republic) (Cornia and Paniccià, 2000). A first increase in differentials was observed by age group, with the death rates of the working age population rising sharply, while those of children and the elderly stagnated or improved. Second, in countries such as Russia and Latvia, the mortality differentials widened also by gender, level of education, employment status, migrant status, and marital status. Shkol- nikov and Cornia (2000) show, for example, that in Russia the life expectancy gap between the best- and least- educated adults rose from 1.63 in 1988-9 to 1.89 in 1993-4.

Pathways of transmission and evidence of impact 21 Globalisation and Health: Pathways of Transmission and Evidence of its Impact

4. Mortality models

hat are the possible explanations for the sation’ from that of endogenous changes or random changes in health indicators outlined above? shocks, such as progress in health technologies or nat- WIn order to answer this we try in this section to iden- ural disasters. It is thus difficult to draw strong conclu- tify the various “pathways” through which changes in sions on causality, and care is required in identifying economic and social determinants can affect health the contributing factors and confounding variables. status. While economic globalisation affects several determinants of health status (see below), the data For these reasons, we review the main mortality mod- presented in Table 4 on mortality by cause of death els which can be used to explain the variations in mor- and population group suggest that the relation be- tality over time and across countries. The first of these tween determinants and health status is strongly non- models emphasizes the material deprivation pathway, linear, and varies considerably across countries.5 Even i.e. the lack of material resources for health, and is most in broadly similar countries, the health impact of a relevant when analysing shifts in mortality due to given economic change may vary due to differences malnutrition, infectious, and sexually transmitted dis- in initial conditions – such as the size of the country, eases, in low income societies among poor households Gross Domestic Product per capita (GDP/c), income or in countries experiencing deep recessions. The sec- distribution, health infrastructure, and domestic poli- ond model emphasizes the role of technical progress cies and institutions. Moreover, as changes tend to in health, and the transfer of health technology across occur at the same time, it is not easy to distinguish countries, as well as the role of the domestic institu- the health impact of ‘policy-driven economic globali- tions needed to absorb such imported technologies.

5 Although the regions in Table 4 are at different stages of the epidemiological transition, cardiovascular mortality remains the leading cause of death in all of them except for sub-Saharan Africa (Gaziano 2007).

22 Globalization and Health Knowledge Network Table 4: Proportional mortality rate by main disease groups and selected causes, 2001 Africa L.America ME-WA Transition SE&E Asia E.Asia/Pacific Advanced hc-ha hc-vha lc-la hc-ha lc-la hc-ha lc-la lc-ha SEA SA vlc-vla lc-la Euro NA 1. Communicable/ Nutr/perinatal (of which) 68.0 73.1 18.5 36.4 17.8 49.3 9.8 4.3 29.2 42.1 11.9 15.0 5.8 6.3 - TBC 3.5 2.9 1.0 3.4 1.1 3.7 1.0 1.4 7.3 4.4 0.5 3.3 … … - AIDS 9.3 28.4 1.3 7.3 … 1.6 … 0.5 2.7 3.1 … 0.5 0.2 0.5 - diarrhoea 6.2 6.8 1.8 4.7 2.7 8.9 0.8 1.0 2.0 6.2 … 0.7 … … - child diseases 9.2 4.6 … … … 5.7 0.3 … 1.8 2.6 … 0.5 … … - malaria 11.3 7.5 … … … 1.6 … … 0.4 0.7 … … … … - mater/perinatal 8.9 6.8 5.2 6.2 4.2 10.2 2.2 0.5 5.2 8.7 0.2 3.6 0.2 0.6 - malnutrition 1.6 1.3 1.5 2.1 0.5 … 0.2 … 1.6 1.2 0.2 0.2 0.2 0.3

2. Degenerative 25.2 20.0 69.1 53.0 67.2 42.1 84.5 83.3 58.1 48.2 80.9 74.5 89.4 87.3 (of which) - cancer 5.5 4.8 15.0 13.6 11.3 5.8 14.9 14.1 10.5 7.2 29.6 17.7 26.0 23.5 - diabetes 0.5 0.6 5.2 2.8 2.3 1.0 1.5 0.6 2.8 1.4 1.5 1.5 2.2 2.8 - cardiovascular 11.0 8.0 29.5 18.5 39.5 22.0 56.4 59.3 26.0 26.3 34.0 32.0 43.2 40.3

3. Violent (of which) 6.8 6.9 12.4 10.5 15.0 8.6 5.7 12.4 12.5 9.7 7.3 10.5 4.8 6.4 - suicide 0.3 0.3 1.0 0.5 1.4 0.7 1.1 2.6 1.6 1.6 3.0 2.7 1.3 1.3 - homicide 1.0 1.1 4.6 2.0 0.9 0.5 0.4 1.6 0.6 0.5 … 0.6 … 0.7 - war 0.5 1.5 0.3 … 0.3 1.6 … 0.3 0.5 … … … … … Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Source: authors’ computation on the basis of WHO data. Note: hc,ha, vha, lc,la, vlc, vla refer to high-child mortality, high-adult mortality, very high adult mortality, low child mortality, low adult mortality, very low child mortality and very low adult mortality. ME-WA = Middle East and West Asia, SEA = South East Asia; NA = North America The third mortality model focuses on the mortality to a rise in mortality due to an increase in infectious impact of acute-psychosocial stress caused by sudden diseases, nutrition problems and violence. In some and unexpected changes and rising uncertainty, and countries, one single model can be used to explain the is most relevant to studies of adult mortality due to observed health trends, but in heterogeneous societ- heart, cardiovascular problems and violence in coun- ies (such as those of Latin America or other middle tries undergoing rapid change and social upheaval. The income countries) two or three of the above mortality fourth model emphasizes the role of unhealthy lifestyles models have to be used to explain mortality changes, such as smoking, drinking, bad diet and lack of ex- as different models have different effects on the health ercise, and is most relevant to mortality changes due status of different social groups. to chronic diseases such as cancer, diabetes and some hearth diseases in affluent societies, but also to an in- 4.1 Material deprivation creasing extent in low and middle income countries. The fifth model emphasizesinequality and lack of social This model is particularly relevant to low- and middle- cohesion as contributing to chronic stress in hierarchical income countries and very poor households in advanced societies or reducing the poor’s access to health care nations. It focuses on the household-level availability of in highly segmented societies. Finally, mortality spikes real resources and the extent of vector and environmental or – more rarely - departures from a long term trend contamination. An increase in real resources (or a reduc- are sometimes explained by random shocks (disasters, tion in contamination) leads to a reduction in mortality famines, wars and epidemics) which generally lead due to infectious, parasitic, airborne, waterborne and

Pathways of transmission and evidence of impact 23 nutritionally-related diseases. For instance, the reduc- Third, instability affects health through its negative tion in mortality observed in Europe during the 19th effect on short and medium term growth. A growing century, well before the introduction of modern medi- body of research shows that volatility has a signifi- cal technologies, was due to improvements in food in- cant effect on investment and medium term growth take, housing, water systems, sanitation and sewage, (Ramey and Ramey 1995). There are several reasons made possible by a rise in personal incomes (McK- for this. Firstly, volatility reduces investment by rais- eown 1976). We now discuss each of the key health ing risk and risk aversion among agents, affecting their determinants in this model: creditworthiness, and pushing up interest rates. A growth in risk aversion may also induce households to (i) The first is real household income per capita often ap- increase precautionary savings and hedge their wealth proximated in aggregate studies by real GDP/c or data through capital flight. In developing countries, vola- on employment and real wages (the latter being less ap- tility in output can affect the stability of public expen- propriate in countries with a large informal sector). This ditures, due to the weakness of automatic stabilizers, variable is particularly relevant in the 100 or so coun- and the high revenue/output elasticity which is typical tries where GDP/c is less than US$4,000. Household for these countries. Volatility in prices (for goods, ex- income often does not include public transfers in kind ports, exchange rates, etc) increases uncertainty. These and cash received through safety net programs. It is changes depress aggregate demand and short term therefore useful to look separately at income transfers such growth, while falls in investments and expenditures as food subsidies and other transfers in kind which can on pro-poor, pro-growth items – such as health and influence the health status of the subsidy recipients and education – may slow down medium term growth. their families. In developing countries, the ‘incidence’ of such subsidies (the ratio of the subsidy to pre-transfer In subsistence societies, the main source of income income) is almost invariably progressive, although the instability are seasonal cycles and weather shocks. In ‘distribution’ (i.e. the share of the total subsidy received globalized societies, instability arises due to fluctua- by each class) may sometimes be regressive. Therefore tions in world prices and in the demand for primary any reduction or elimination of these subsidies may af- commodities (Guillaumont et al. 2006), but also in fect the health of those most in need. interest rates, borrowing conditions and capital flows, particularly short term capital flows. When export (ii) Instability of household income. Income is generally earnings are highly volatile, governments experience assessed over a given reference period (usually one large revenue fluctuations, making it difficult for them year). The ability to secure food, shelter and medicine to manage external debt and sustain health expendi- over this period is based on the assumption that the tures. Farmers, traders and local trade or financing in- household is able to spread its income over the entire stitutions tend to have irrational expectations about period considered, regardless of the exact period in future prices in boom periods, and on the other hand which the income is generated. This implies that the show strong risk aversion in bad years. As a result there household can borrow in periods of low earnings, and is reduced investment and use of inputs. save in those of high earnings. But such ‘consumption smoothing’ is not feasible where financial markets (iii) Income Inequality. In aggregate studies, the impact of are absent, costly or imperfect, as is the case in many household incomes on health is approximated by an developing countries. Households may therefore face index of income inequality. Given an average GDP/c, a health-threatening fall in consumption, with irre- an egalitarian distribution can improve health status versible health effects, during periods of low earnings, by ensuring that most households have enough in- even if their average per capita income for the year is come to satisfy their basic needs, while high inequality above the poverty threshold. Second, instability affects limits the access of the poor to basic resources. health negatively because it leads to a rise in income inequality. Indeed, the evidence suggests that the poor High inequality also affects future health status be- are more affected by instability than the rich, who ex- cause of its negative effect on the growth rate of GDP. perience a less than proportional decline in their in- Post-Keynesian inequality theories which focus on come (Guillaumont et al. 2006). differences in rates of savings of profit recipients and

24 Globalization and Health Knowledge Network high-wage earners in relation to low-wage earners (Pa- other basic traded items rises faster than the average sinetti 1974) argue that high initial inequality favours price level (the same effect has been recently observed long term growth. Similar conclusions have been in the EU countries). For instance, much of the re- made in other theories used to analyze social mobil- cent analysis of rural poverty in India shows that the ity in unequal societies. Yet, most theoretical analyses CPI of agricultural labourers has risen faster than aver- and empirical studies show that inequality correlates age prices and agricultural wages (Sen and Himansu negatively with GDP growth. The ‘political economy 2004). A similar effect occurred in the 1990s in the models’ (Alesina and Rodrik 1994) argue that high transition economies where price liberalisation led to initial inequality damages growth as it leads to the a faster rise in the prices of those basic goods which election of governments which favour redistribution were heavily subsidized during the socialist era. A pos- through high marginal tax rates, which depress private sible way to allow for this effect in regression analysis investment and growth. In ‘capital-market imperfec- is to include the ratio of the index of food prices to tions’ models (Aghion et al. 1999) high inequality CPI in the explanatory (right-hand side) variables. harms growth as it leads to slow human capital formation, locks investments by the rich in low return (v) The level of education of family members, mothers in activities while the poor - who tend to have projects particular, is a major determinant of the health sta- with higher rates of return - cannot invest more than tus of all family members, but especially of children. their limited endowments due to capital market im- Educated parents not only tend to have better labour perfections. A third strand of the literature (Venieris market and fertility outcomes, but they also manage and Gupta 1986, Bourguignon 1998) shows that high household resources more efficiently, make more ra- inequality may cause street protests, violent rioting, tional consumption decisions for themselves and their high crime and other actions which create uncertainty children, make better use of public health services, among investors, erode property rights, raise transac- and are more aware of their rights when dealing with tion and security costs and reduce growth. Fourthly, health providers. Education also facilitates improve- high asset and income inequality reduce the scope for ments in health knowledge, practices and behaviour conducting rational economic policies, as they restrict (e.g., washing hands before preparing food, observing the supply of pro-growth public goods (such as law other aspects of personal hygiene, and avoiding un- and order, contract enforcement, security of property healthy lifestyles including unsafe sex), and empowers rights, and human capital), lead to the adoption of lax parents in their function as health providers. In addi- macroeconomic policies and a high risk of defaulting tion, education facilitates the process of acquiring and on international debt, as governments are unable to processing new types of information, which can be a tax the elites and are so impelled to borrow abroad positive advantage during difficult periods of transi- (Birdsall 2000). Finally, Cornia et al. (2005) argue tion and social upheaval. that, beyond a given threshold, inequality affects microeconomic incentives, increases labor shirking, free All empirical studies of developed and developing riding and supervision costs, erodes the social con- countries have found that the level of education of tract, and may force the poor to over-exploit common the mother is a major predictor of IMR and U5MR goods such as forests and grazing land. (Caldwell 1979, Bicego and Boerma 1993). The impact is, however, clearly non-linear. It is not very pro- (iv) Relative prices of essential goods. The income (or GDP/c) nounced for women with elementary education, but variable discussed above is expressed in real terms, i.e. increases sharply from primary to secondary educa- after adjusting nominal incomes using the GDP price tion and above. The level of education has also been deflator or the consumer price index (CPI). Yet, es- found to be a major predictor of adult mortality in in- pecially in countries affected by high inflation, such dustrialized countries. In 1979-82, for instance, death an adjustment is often insufficient, as the prices for rates for 30-34 year old Czech males with basic educa- basic necessities (food and medicine) fluctuate more tion were 2.9 and 5.7 times higher than for males of strongly than the CPI. Indeed, empirical evidence the same age with secondary and university education shows that in open economies with weak currencies (Hertzman 1995). The same result is obtained even and a rigid agricultural supply, the price of food and after controlling for other confounding variables.

Pathways of transmission and evidence of impact 25 (vi) Access to essential public or private health services is also important resource for health. The amount of time re- an important determinant of health status. Yet, such quired for such tasks varies with the level of GDP/c, as information is rarely available at the household level well as the availability of time-saving household tech- (except in DHS and LSMS-type surveys). Moreover, nologies for cooking, washing, preserving food and so while there is ample micro-evidence that access to on. Time use depends in an important way also on health services and the number and quality of health the demographic structure of the family. Evidence for workers are closely associated with health coverage and both developing and industrialized countries shows health outcomes (WHO 2006), the variables used to that families with unfavorable dependency ratios tend to approximate such phenomena in macro studies (such have worse health outcomes, particularly female-head- as aggregate public health expenditure per capita, ed households or families with many dependents. The public health expenditure as a share of GDP, and the time adults dedicate to family reproduction, as well number of health professional or hospitals/clinics per as the traditional division of labour (production for 100.000 people) are often non-significant in regres- the market by men, household production and family sion analysis. In fact, the allocation of public health reproduction by women), can also be influenced by expenditure depends on complex budgetary processes shocks which act to depress the wages of household and political economies, and is often characterized by heads and other family members engaged in market a sub-optimal allocation of resources between levels of production. This generally leads to an increase in the care and across regions. For instance, the same health time dedicated by women to market production, and resources can be allocated to low-cost, high-impact a reduction in the time assigned to the care of the fam- vaccination campaigns and essential drugs for the ily and to rest. poor, or to urban hospitals serving the middle-upper classes in few cities. While such alternative allocations Unfortunately, there are very few time budget stud- of public expenditure have considerable effects on ies that provide information on the time use of health status, they are not easily captured in aggregate household members. The situation is even worse in data on expenditure and staffing levels. aggregate studies where information on time use is unavailable. One possibility is to proxy this phenom- The type of health financing also affects access to enon by using dependency ratios, or a sudden rise health care and health status. Free public provision in the labour market participation rates of women, of basic health services is the least discriminatory, al- elderly and youth. though the quality of such services often leaves a lot to be desired. This advantage remains, even if ‘nominal’ (viii) Reproductive behaviour. Micro studies have shown that user fees are introduced to reduce wastage and service- the risk of death of children and mothers is strongly overuse without excluding the poor, but disappears correlated with the birth order of children and the when ‘substantial‘ user fees are introduced, since these age of the mother. Very young mothers, particularly usually lead to low income households being excluded if they are single, tend to be neither economically from health care. Private health financing is the most independent nor physically developed or psycho- exclusionary approach, as observed today in countries logically mature enough to bear, raise and educate a such as the USA and China. The impact of health in- child. Children of young mothers, especially when surance varies considerably but can be regressive in unplanned and unwanted, run a far greater risk of economies with a large informal sector. Local-level fi- accident, abandonment, institutionalisation, poverty nancing and community-based health care provision and psychological maladjustment than the children of can replace eroding state bureaucracies in the supply older and better educated parents. Likewise, high par- of basic health services, but the effects are not always ity births are more likely to occur to women who are positive. least able to support healthy pregnancies and thus face greater delivery problems. For these reasons, infant, (vii) Family characteristics.The time available to family child and maternal mortality correlate closely with to- adults, women in particular, for the reproduction of tal fertility rates, as – ceteris paribus - a decline in the the labour force, i.e. for food preparation, cleaning, latter automatically entails a decline in high-parity, and care of children, the elderly and sick is also an high-risk births for very young and old mothers.

26 Globalization and Health Knowledge Network (ix) Finally, vector and environmental contamination af- The above authors foresee no reduction in the flow fect life expectancy, particularly among the poor, of new inventions. They note however that the health who are more likely to live in less sanitary areas, benefits they generate will mainly accrue to high in- and therefore be exposed to several air-, water- and come countries and people, with the risk of increasing vector-borne diseases, various pollutants, hazardous differentials in mortality rates between countries, at industrial wastes, and pesticides. The spread of these least in the short term, and that the ability to con- diseases varies according to economic conditions trol such differentials depends crucially on the cost of and the prevalence of preventative public health transferring health technologies. measures, and therefore tends to slow down with a rise in GDP/c. Indeed, with the exception of typhus Measuring progress in medical technologies is not and cholera, vector and viral contamination has been easy. To the best of our knowledge, there are no syn- steadily brought under control over the last decades. thetic indexes which capture the myriad of new dis- However, malaria remains a largely unresolved prob- coveries. In addition, the health impact of new discov- lem. In addition, the assumption of the early 1970s eries depends on their accessibility to the population that infectious diseases could be progressively elimi- at large, and this depends in turn on their price and nated, cannot be shared today after the appearance the existence of appropriate delivery channels. A pos- of new diseases such as hepatitis D, Ebola and other sible way to measure progress in health technologies haemorrhagic fevers, AIDS, tularaemia, lyme disease in regression analysis is to use linear, convex or con- and Lassa fever (Livi-Bacci 2000). Meanwhile emis- cave time trends, or period dummies. Such variables, sions of CO2 and other pollutants have increased, however, may also capture the effect of other unob- and the growing international trade in toxic waste served phenomena. A second way to proxy progress in lead to an increase in health risks for people in some health technologies in developing countries is to use areas of low income countries. data on the coverage rates of specific programs such as immunisation, control of malaria-AIDS-TB, oral re- 4.2 Progress in health technology hydration, prenatal care, clinic-based delivery, breast- feeding and access to fresh water supply. In advanced According to this model, progress in medical technol- countries with a large share of elderly people, the most ogy and health knowledge is the most important fac- common proxy is the percentage screened for chronic tor influencing declines in mortality, and one which and degenerative diseases such as diabetes, cancer, and has a greater influence tthan the material welfare of high cholesterol. Such variables are highly significant households. In the post World War II period, progress in regression analysis. in this area contributed to the reduction in mortality due to infectious, parasitic and communicable diseas- 4.3 Acute psychosocial stress es through the introduction of low-cost, high-impact preventative measures, such as malaria spraying, vac- Stress is a key factor in deaths due to heart problems, cination with freeze-dried serum, distribution of anti- hypertension, alcohol psychosis, neurosis, suicide and biotics, and oral rehydration therapy (Preston 1976). accidental deaths, ulcers and cirrhosis of the liver. Ac- Many developing countries which recorded slow or cording to Serafino (1994, cited in Shapiro 1994): little economic growth experienced a substantial reduction in mortality due to the introduction of new Stress is the condition that results when person/ health technologies. And in advanced societies, prog- environment transactions lead the individual ress in medical knowledge made it possible to achieve to perceive a discrepancy – whether real or a sharp decline in cardiovascular and cancer-related not – between the demands of the situation mortality, especially among the middle-aged and el- and resources of the person’s biological, derly population. Cutler, Deaton and Lleras-Muney psychological or social systems. (2006) suggest, for example, that reductions in cardiovascular mortality due to medical advances explain In countries hit by social upheavals, going through pro- 70 percent of the seven year gain in life expectancy found transformations, or confronted with the effects recorded in the USA between 1960 and 2000. of globalisation, deaths due to acute psychosocial stress

Pathways of transmission and evidence of impact 27 reflect an increase in pressures to adapt to new and un- found that - after controlling for education, income, expected situations, the absence of strategies for coping age, gender and marital status - unemployed people with them, and the inability of public policy to respond were between 22 and 41 percent more likely to suffer to these challenges. Epidemiological research has shown from bad health than those who were employed. that acute stress leads to physiological and psychological arousal, which provokes sudden changes in heart rate, Drawing on the existing literature, Marmot and blood pressure and viscosity, a reduction in the ability Bobak (2000) suggest that jobless workers face on av- to maintain emotional balance, and a pervasive sense of erage a 20 per cent greater risk of mortality than the uncertainty, powerlessness, and loss of social role, per- employed. In addition to loss of income, unemploy- sonal control and purpose in life. Acute psychosocial ment contributes to a loss of skills, cognitive abilities, stress has been shown to lead to increased consumption motivation, sense of control and self-respect, and feel- of health-damaging ‘stress-relievers’ such as alcohol and ings of being unwanted, unproductive, dependent, drugs, which further affect mental balance and social and without a social role (Sen 1997). For young peo- behaviour. Stress affects men more than women, who ple, unemployment and unstable employment are an are protected against stress by a lower production of tes- obstacle to marriage and a stable life. Unemployment tosterone and who, unlike men, can generally draw on may also erode social norms and lead to an increase in a more diversified range of activities and social relations crime rate among the jobless, disrupt social relations to shelter them during difficult periods. Yet, this rule and can be associated with an increase in cases of fam- of thumb does not always apply: for instance, during ily violence. the Chinese transition, stress-related suicides rose more rapidly among women (Lu, 2006). Other sources of job-related stress include involun- tary rapid job turnover, informal and insecure employ- It is difficult, if not impossible, to provide a list of all ment, arbitrary reassignment to stressful occupations, the possible sources of stress, as some shocks may be and enterprise restructuring. Similar effects have been better controlled in some societies than others, thanks observed among managers from bankrupt firms who to better institutions or greater innate resilience. How- feel ashamed of their perceived inability to save the ever the Social Readjustment Classification of stress firm and worry about the social consequences of their suggests that the impact of some stress factors has been firm’s failure, individuals who have incurred large -fi noted in many places, especially when the formal and nancial losses, and small producers affected bydeterio - informal institutions and coping strategies needed to rations in inputs/output price leading to immiseration limit their impact are absent or no longer operational. and indebtedness. Various surveys of the poor have The main sources of stress can be classified as follows: shown that high inflation, uncertainty about relative prices and erosion of real financial assets are also key (i) A first type of stress may arise from the work sphere. A stress factors. In addition, high inflation rates can af- key stress factor is loss of employment, especially if un- fect health in a variety of ways. It erodes real income anticipated and unaddressed by public policy. Studies and assets for those who cannot dollarize, causes a of factory closures and follow-up studies of unemploy- fiscal drag, and raises anxiety about the future. High ment in developed and transition countries found evi- inflation may also discourage investment and entre- dence of short-term adverse effects on mental health, preneurship because of the uncertainties it creates. suicide, ulcer prevalence, and diseases related to the Overall economic and political instability is also a stress circulatory system (see the literature in Cornia and factor, due to the economic responses it elicits (see Paniccià 2000). Other studies have focused on the later), and the all-pervasive ‘fear of the future’ which relation between ‘health duration’ (i.e. the length of it generates. time an individual enjoys good health) and unemployment. Cooper et al (2006) use such a longitudinal (ii) Stress may also arise also from the sense of frustration methodology (which can control for reverse causation caused by failure to fulfil one’s family role as a result and endogeneity problems), and draw on the Physical of employment loss, wage declines or other factors. In and Mental Health Problems, Illnesses and Disabil- the patriarchal middle-income societies of Eastern Eu- ity measure of health for 13 European countries. They rope or Latin America, a sudden wage decline may lead

28 Globalization and Health Knowledge Network to an increase in mortality among household heads who remain in their community. As noted by Emile not because of shortages of basic goods, but due to Durkheim (cited in De Vogli 2004), distress migra- their perceived inability to provide their families with tion is a source of ill health, as it entails considerable those goods perceived as ‘socially necessary’. Shame material hardship, disorientation and loss of control and frustration are more likely to be experienced by in new environments, breakdown of family and so- men (as the family’s main provider) than women, de- cial relationships, housing problems, social exclusion spite the latter’s high rate of participation in the la- and discrimination, a redefinition of survival strate- bour force. gies and, as a result of all this, greater stress and risk of death. Finally, personal insecurity and fear, such as (iii) A third set of stress factors are rapid changes in social those experienced in high crime rate areas, are also a hierarchies and roles, a commonly observed phenom- major source of stress. enon during major upheavals such as political changes, radical economic reforms, and rural-urban migration. 4.4 Unhealthy lifestyles Rapid change and the ensuing social re-ranking, mean for some a loss of role, reputation, and influence. If the This mortality model emphasizes the role of health rise of new elites is perceived to be due to ascription and knowledge and practices and personal behaviour in corruption, rather than hard work and competence, the health improvements. In this model, the promotion of losers experience frustration and rage for what is per- health information and changes in personal behaviour ceived undeserved demotion, envy, hopelessness, and are more important long-term determinants of health loss of purpose. This was common among the middle- status than income, stress or health technology. Smok- aged labour elite of semi-skilled industrial workers and ing, drinking, lack of physical exercise, and an unbal- communist cadres in the former socialist countries, and anced diet have been shown to be key risk factors for among the rural elders in societies migrating to cities. several illnesses. Smoking is a major factor contribut- For these groups, social disorientation is acute as the ing to premature death from cancer of the lung, blad- values, norms and roles of their lifetime are replaced der, mouth and stomach, emphysema, cardiovascular by new norms, and their ability to adjust to new con- diseases, cirrhosis, and non-medical fatalities (Figure ditions, grasp retraining opportunities and social pros- 1). Non-smokers are also affected by cigarette smoke – pects is usually limited. Since changes in hierarchies with increased risks of lung cancer and cardiovascular and roles are not easily observable, they are often ap- disease (World Bank 1999), while women who smoke proximated by indexes of social mobility or large shifts while pregnant are more likely to have stillborn or low in income distribution; or by a rise in inflation, which is birth weight babies. Smoking currently kills one out of an excellent index of redistribution of income, wealth every ten people. It has been estimated that if current and power from low income wage earners and pension- trends continue, smoking will be the leading cause of ers to the upper classes who are more likely to own real death in the world by 2030 (World Bank 1999). Today, assets or hard currency. about 1.1 billion people smoke, of whom 85 per cent live in developing countries (Ezzati et al. 2005) where (iv) A final source of stress concernspersonal living arrange- quit-rates are very low (2 percent in India, 5 percent in ments. At all ages, stress-related mortality is signifi- China, as opposed to more than 30 percent in the UK), cantly higher for people who are widowed, divorced and where nearly one third of all tobacco users die from and single than for those who are married. The latter cardiovascular diseases (Gaziano 2007). supposedly lead healthier lifestyles, are less exposed to stress and have greater access to social support net- Excessive consumption of salt and saturated fats in- works than people living in incomplete families. Ruz- creases the incidence of cardiovascular problems, icka et al. (1989) suggests that the increase in suicides while low intake of antioxidants is associated with a observed in the developed countries during the 1970s higher risk of myocardial infarction (Bobak and Mar- and 1980s may have been partly due to a decline in mot 1996). Excessive alcohol consumption is related marriage rates and a rise in divorce rates. Past analyses to deaths due to cirrhosis of the liver, alcohol psycho- have shown that migrants – regardless of their family sis, accidents and injury. The incidence of these risks arrangements - face greater mortality risks than people correlates weakly with the level of income and health

Pathways of transmission and evidence of impact 29 expenditure, but is inversely related to the level of edu- While there are numerous clinical studies measuring the cation. Such variables affect health status gradually as health impact of unhealthy lifestyles, there are only a a result of the cumulative exposure to risk, meaning few analyses estimating their overall health impact. Part that there are long lags between protracted exposure to of the problem is the lack of nationwide statistics in this unhealthy behaviours and increases in mortality. Only area. Aggregate routine data on alcohol consumption deaths from accidents respond rapidly to an increase are notoriously deficient, and data on physical activity in alcohol and drugs consumption. Changes in life- and diet composition are difficult to obtain in broad styles also tend to occur slowly. surveys. Likewise, information on aggregate smoking prevalence for developing countries is very limited. For years, the unhealthy-lifestyle model has guided mortality analyses in developed countries, but the 4.5 Income inequality, hierarchy recent literature on the ‘nutritional transition’ and and social disintegration ‘communicability of smoking to developing countries’ suggests that it also applies to the urban population It was argued in section 4.1. that inequality raises of middle and some low income countries. However, mortality by reducing access of the poor to food and in very poor households in developing countries, the to private or subsidized health care, while section 4.2. main problem remains an insufficient – rather than noted that high inequality affects health by raising the inappropriate – calorie and protein intake for a size- crime rate, stress and the number of violent deaths. able share of the population. There is also evidence that in advanced countries high

Figure 1: Relation between smoking incidence among 25-55 years old males and LEB

30 Globalization and Health Knowledge Network inequality may affect health status by creating highly inequality and health in developed countries, that stratified and less cohesive societies (Wilkinson 1996 chronic stress due to relative deprivation has a large and Wilkinson 2000).6 The theory of collective action impact on health. The authors base their conclusions suggests that unequal and segmented communities are on evidence regarding the socioeconomic gradient, characterized by strong heterogeneity, divergence in which does not just distinguish between the poor and social objectives, and a limited capacity to undertake the rest of the society, but every rung of the social lad- collective action, including in health care. As econom- der. Finally, high inequality may lead to a hierarchical ic and social disparities widen, the divergence in the organisation of work and society in which most of the interests of social groups increases, taxation and the decision making is concentrated in the hands of a few provision of health services decline, residential segre- people, while the rest is engaged in basic simple tasks. gation rises and political participation and the efficacy People at the bottom of the social ladder generally have of government institutions deteriorate. Such phenom- much less control over work and other life decisions ena are not limited to industrial nations such as the compared to people in higher social positions and, in US and the UK (where they have been the object of extreme cases, are affected by feelings of “learned help- serious research efforts), but are also observed in poor lessness”, a response to a social environment that is un- rural settings. For instance, a study by Godoy et al. predictable and uncontrollable. (2006) on 13 villages of the Bolivian Amazon found that village level income inequality was associated 4.6 Random shocks with anger, fear and sadness, even after conditioning for social capital (which was also negatively affected by The five pathways discussed above do not explain rises inequality), and that all this had a negative effect on in mortality - or major deviations in its trend - due to health outcomes. natural disasters, conflicts, sanctions and epidemics. In this paper, these are treated as random events escap- Secondly, high inequality erodes social cohesion, an es- ing the control of the policy maker, but which have to sential factor for promoting the diffusion of health in- be controlled for in a regression analysis which aims formation, exerting control over deviant health-related to establish a link between economic globalisation behaviour and criminal activity, and providing inter- and health status. Claims that these shocks are due to personal help among community members (Kawachi unsustainable development policies are plausible, but et al, 1999). An example of the positive health impact have not yet been demonstrated conclusively. of strong social cohesion is the well known nine year follow-up study of Alameda County, California (cited (i) Natural disasters and famines. Large and sudden in- by De Vogli (2004)). The study showed that the resi- creases in mortality have at times been caused by dents who were more socially integrated at the outset droughts, floods and other disasters. 10-12 million of the study lived longer than their counterparts who people were wiped out during the Great Leap Forward had fewer social contacts (Berkman and Syme, 1979). of 1958-63 in China, and an estimated two million According to House et al. (1988), the health risk as- during the North Korean famine of the mid 1990s. In sociated with low levels of social integration in highly all these cases mortality due to nutrition-related and unequal societies is comparable in magnitude to the infectious diseases increased mainly among children risks associated with smoking, high blood pressure, and the elderly, but returned quickly to its baseline and obesity and remains significant after controlling value after the shocks disappeared. In many cases, the for these and other risk factors. initial random shock is aggravated by the absence of adequate policy responses. This phenomenon can be Thirdly, it is argued by Wilkinson and Pickett (2006), included in the regression by a dummy variable, or by who reviewed 168 analyses of the relation between the number of deaths caused by disasters.

6 This approach remains however controversial, and some authors (Deaton, 2001) suggest that the findings of these studies should be treated with scepticism, in part because of theoretical problems (i.e. the impact of omitted variables that correlate with inequality), and in part because of the weakness of the empirical evidence provided. Deaton emphasizes in particular the limited reliability of the inequality measures used, especially for international comparisons. The author does not exclude the possibility that inequality is one of a multitude of factors affecting health, but rejects the view that inequality is – as argued by the supporters of this viewpoint – the main determinant of ill health.

Pathways of transmission and evidence of impact 31 (ii) Conflicts and complex humanitarian emergencies. ing decline in population size or in its rate of growth. Deaths due to violent causes, starvation and infectious AIDS is taking on the characteristics of a pandemic diseases increase during wars, domestic conflicts and – rather than an epidemic – which might gradually ethnic struggles. Such deaths often result from the uplift the mortality curve. massive exodus of refugees and internally displaced people brought on by the conflicts. As in the previous 4.7 Summary of variables affecting mortality case, a sudden rise in death rates is usually followed by by the five main pathways a return to pre-crisis levels in just 1-2 years after the end of the shock. However this may not be the case in Table 5 summarizes the main determinants of health parts of sub-Saharan Africa, where internal conflicts by type of pathway. have become endemic, and problems due to displacement of refugees and distress migration take a long time to resolve.

(iii) Sanctions. Economic sanctions can often have unexpected effects on mortality. This may be due to the -de clining availability of essential imports, which affects not so much the elites but the poor and biologically vulnerable. For instance, while the 1961 US embargo against Cuba initially had a limited health impact because of the assistance provided by the Soviet Union and the egalitarian policies in the allocation of scarce food and medicines, the 1989-93 recession caused by the collapse of the Soviet Union and the tightening in 1992 of the US embargo – that affected the availability of nutrients, water sterilization products, drugs and vehicle spare parts – had a strong health impact on Cubans. Mortality per 1000 inhabitants rose from 6.4 to 7.2 in 1989-94, and caused some 7500 excess deaths due to TB, infectious and parasitic diseases, diarrhoea, influenza and pneumonia among the population over 65 years of age. Garfield and Santana (1997, quoted in Cornia 2004) note that while some of these deaths were caused by the recession of 1989-92, the tightening of the US embargo aggravated the situation. A much greater number of additional deaths were recorded, in turn, by the imposition of sanctions on Iraq during the 1990s.

(iv) Epidemics. Large mortality increases may occur due to the spread of previously unknown diseases, as in the case of the black plague in Central Europe in the fif- teenth century, virulent influenza in 1918, and HIV/ AIDS in the late 1980s. Such epidemics do not have a similar time profile. Some of them generate their effects quickly, while others (such as HIV) have a long period of incubation and their effects are thus lagged. Also, some epidemics become endemic, and their impact can extend over several decades, leading to a last-

32 Globalization and Health Knowledge Network Table 5: Summary of variables affecting health status by main mortality model Material Technical Psychosocial- Unhealthy Social cohesion deprivation progress stress lifestyles 1. Income/c XXX ------Δ income/c ------XXX ------income instability XXX --- XXX ------Income transfers XXX ------XXX 2. Gini income XXX --- XXX --- XXX Δ Gini income > 4 ------XXX ------3. Unemployment rate XXX --- XXX --- XXX Job turnover ------XXX --- XXX Large fall in wage rate ------XXX ------Bankruptcies ------XXX ------4. Inflation rate ------XXX ------Relative price of food XXX --- XXX ------5. Female Illiteracy XXX XXX --- XXX --- Women with 1ary educ.. XXX XXX --- XXX --- Women with 2ary educ. XXX XXX --- XXX --- 6. Health expenditure/c XXX XXX ------Doctors/1.000 XXX XXX ------Out of pocket expenditure XXX XXX ------XXX 7. Techn. Progress health --- XXX ------Coverage of PHC progr --- XXX Screening of degen disea. --- XXX ------8. Time use XXX ------XXX --- Dependency ratio XXX ------XXX --- 9. % incomplete families XXX ------XXX --- Total Fertility Rate XXX ------Migrants/pop ratio XXX --- XXX ------10. Environ. contamination XXX --- XXX ------11. Alcohol consumption ------XXX XXX --- Smoking incidence ------XXX XXX --- Unbalanced diet ------XXX --- 12. Natural disasters XXX --- XXX ------War and conflicts XXX --- XXX ------Sanctions XXX --- XXX ------Epidemics (HIV-AIDS) XXX --- XXX ------

Source: authors’ compilation, Notes: XXX indicates that the variable is relevant, --- that it is not.

Pathways of transmission and evidence of impact 33 Globalisation and Health: Pathways of Transmission and Evidence of Impact

5. Changes in the social determinants of health:1980-2005 compared to 1960-80

lesser extent – the Asian economies affected by the fi- his section discusses the changes in the determi- nancial, banking and currency crises of 1997-8 which nants of health summarized in Table 5 and their depressed world growth by one percentage GDP Tlink to liberalisation and globalisation policies. point.

5.1 Household Income This growth slowdown was somehow unexpected, although – as shown in Table 6 – some deceleration Changes in this area are well documented (Rodrik 1999, was already evident in the 1970s in Eastern Europe, Berry and Serieux 2006) and are only briefly summa- South Asia, Sub-Saharan Africa and the OECD coun- rized here. Four main trends are identified in the cur- tries. While the 1990s saw a rise in the number of rent debate and are illustrated in Tables 6 and 7: civil conflicts and natural disasters (see below), these had a limited effect on growth, as they generally af- (i) The growth of the world economy slowed down over fected small countries. At the same time, as noted in the 1980-2000 period (‘the era of the Second Globali- the introduction, the decade benefited from a number sation’) compared to 1960-1980 (the ‘Second Golden of positive dividends, as well as from historically low Age of Capitalism’). The annual average rate of growth real interest rates on dollars loans that – ceteris paribus of world GDP per capita fell from 2.54 per cent in – should have raised rates of growth and household 1960-80 to 1.39 in 1980-2000. Growth was particu- incomes. Despite all this, the world rate of growth of larly weak in the 1990s due to stagnation in Europe GDP/c barely changed as the moderate improvement and Japan, the collapse of the European countries in recorded in middle income countries was counterbal- transition, the persistent difficulties faced by Latin anced by the sharp recessions recorded in the tran- America and Sub-Saharan Africa despite widespread sition economies, persistent stagnation in Africa and policy reform, and stagnation in MENA and – to a further economic decline in the advanced economies.

34 Globalization and Health Knowledge Network Table 6: Period GDP/c growth rates* by main regions, 1960-2005

1960- 1970- 1980- 1990- 1960- 1980- 2000- 1970 1980 1990 2000 1980 2000 2005 High income countries 4.11 2.60 2.39 1.88 3.35 2.13 1.45 China 1.49 4.31 7.71 9.26 2.89 8.48 8.77 East Asia & Pacific (excl. China & Japan) 2.87 4.51 3.47 3.18 3.69 3.32 3.60 Eastern Europe & Central Asia 5.0* 2.3* 2.1* -1.03 n.a n.a 5.07 Latin America & Caribbean 2.54 3.15 -0.89 1.68 2.85 0.39 0.94 Middle East & North Africa n.a n.a 0.14 1.88 n.a 1.01 2.15 India 1.69 0.68 3.58 3.62 1.19 3.60 5.25 South Asia (excl. India) 2.60 0.40 2.82 1.93 1.50 2.38 2.52 Sub-Saharan Africa 2.31 0.76 -1.04 -0.32 1.53 -0.68 1.96 World 3.25 1.83 1.38 1.41 2.54 1.39 1.53

Source: authors’ calculation based on WDI (2006).Notes: the regional aggregates refer only to the developing countries (e.g. East Asia does not include Japan); The data in the columns refer for the periods 1960-70, 1970-80, 1980-90 are from Cornia and Danziger (1997); * average yearly compounded and population weighted growth rates computed on a point-to-point decennial basis.

Table 7: Share of countries (by region) which experienced negative growth of GDP/c

1960-1970 1970-1980 1980-1990 1990-2000 2000-2005 High income countries 0.0 0.0 0.0 0.0 8.7 East Asia 0.0 0.0 36.4 8.3 0.0 Latin America and the Carribean 0.0 20.8 48.0 28.0 16.0 MENA 0.0 0.0 37.5 10.0 10.0 OECD 0.0 0.0 0.0 4.2 4.0 South Asia 0.0 20.0 0.0 0.0 0.0 Sub-Saharan Africa 28.0 27.6 56.2 48.6 31.4 European transition economies 0.0 0.0 14.3 28.6 0.0 USSR 0.0 0.0 6.7 93.3 0.0

% of total countries with growth rate of GDP/c < 0 6.3 11.8 30.7 32.1 12.5 % of total countries with growth rate of GDP/c < 1 15.3 18.5 48.7 44.8 30.5

Source: authors’ calculation based on WDI (2006).

The data on the share of countries experiencing a neg- growth of less than 1% in another 12-18 percent of ative rate of growth of GDP/c (Table 7) confirm the the countries, particularly the low-middle income trends illustrated in Table 6. nations of Eastern Europe, MENA, Sub-Saharan Af- rica, and Latin America (Table 7). In the latter region, (ii) The unweighted standard deviation of the growth which faithfully implemented Washington Consensus rate of GDP/c increased (Figure 2). Indeed, nega- (WC) policies, only three countries recorded higher tive or zero growth was recorded in 32 percent of rates of growth in the 1990s than those recorded in the countries analyzed over the last two decades, and the 1950-1980 period, which was dominated by

Pathways of transmission and evidence of impact 35 F igure 2.. S tandard deviation of G DP Figure 2: Standardgro deviationwth ra ofte GDPs ,, 1 growth950 - rates,200 11950-2001

4.0 1 4 2 c o u n trie s D e v e lo p in g c o u n trie s

3.5

3.0

2.5

GDP growth 2.0 GDP growth 1.5 Standard deviation of per capita Standard 1.0 Standard deviation of per capita 1950-1973 1973-1980 1980-1990 1990-1997 1997-2001

inward looking policies. With this slowdown, health (iv) During the last five years the growth rate of world GDP status – ceteris paribus - improved more slowly than per capita showed signs of recovery due to the grow- in the past (see below). In contrast, despite growing ing weight of the Asian emerging economies, which distributive tensions, growth improved in East and now account for close to half the world GDP-PPP, South East Asia, India and – especially – China. In and which have contributed to a recovery in sub-Saha- the 1978-2003 period, this latter country doubled its ran Africa, the resumption of growth in the European already respectable rate of growth for the 1970s, and economies in transition, and the moderate recovery of quintupled its GDP/c. Such an acceleration of growth the Japanese economy. This moderate growth upsurge seems likely, ceteris paribus, to have generated gains in was however offset in part by prolonged stagnation of health status, which were however offset by a rise in the main EU economies. income inequality, adverse changes in health financing and rising psychosocial stress (Figure 2). 5.2 Income subsidies

(iii) The engine of world growth shifted from the OECD There are no ready-made compilations of income to East and South Asia. However, given their small or consumption subsidies (except for the highly ag- initial GDP, the growth upsurge in these countries gregated IMF-GFS data presented in Table 8). It is was more than offset by the slowdown in the OECD therefore difficult to assess trends in this area. Gener- countries, leading in this way to the global decelera- ally speaking, in the 1980s and 1990s public policy tion mentioned above. As noted in the introduction, emphasized the reduction of subsidies, while pro- the acceleration in the tempo of growth recorded in moting targeted ‘unconditional transfers’ (such as China, India and the rest of the region was the result the Child Support Grant in South Africa) and ‘con- of the introduction of home-made selective and grad- ditional transfers’ (such as the Brazilian Bolsa Escola ual policy reforms - including some domestic liberali- and the Mexican Progresa). Such transfers provide a sation, export orientation – which differed markedly cash reward to the families of children who attend from the WC and Augmented Washington Consensus health check-ups and school regularly. The data pre- (AWC) blueprint. sented in Table 8 suggest that there has been a de-

36 Globalization and Health Knowledge Network Table 8: Public subsidies / GDP, selected years

1970 1975 1980 1985 1990 1995 2000 High income countries 0.75 1.03 3.34 2.48 1.59 1.49 1.81 Middle income countries … … 0.59 1.97 1.51 0.76 0.75 Low income countries … … 1.22 0.83 0.88 0.83 0.95 Transition economies … … … 15.44 7.96 2.35 2.09

Source: authors’ elaboration on IMF-GFS data.

Table 9: Percentage of households in each income decile receiving different subsidies, Viet Nam 1993 and 1998

Social pensions, war veterans Production and Poor relief pensions, unemployment benefits consumption subsidies (various programs) 1993 1998 1993 1998 1998 decile1 (poorest) 6.25 2.92 13.13 7.85 8.31 decile2 11.25 5.89 11.67 8.98 2.21 decile3 11.04 7.60 9.58 8.36 2.13 decile4 11.04 9.05 10.83 10.14 1.72 decile5 15.42 9.62 10.00 11.58 1.31 decile6 17.08 14.12 9.38 11.53 1.95 decile7 14.79 17.01 10.42 10.24 1.22 decile8 22.29 16.70 8.54 9.69 0.90 decile9 16.25 21.88 8.33 9.21 0.38 decile10 (richest) 17.95 22.13 11.06 8.81 0.41

Source: Tiberti (2006) on the Viet Namese Living Standard Measurement Surveys 1993 and 1998.

cline in the ratio of production and consumption 5.3 Economic instability subsidies to GDP, but these highly aggregated data should be treated with caution as they represent the In high income countries, macro instability remained average trend observed in a limited (and changing) broadly unchanged or even declined. The same cannot number of countries. On the other hand, there is evi- be said about micro instability, as has been shown by a dence that the volume of transfers declined in transi- myriad of mobility studies. In contrast, macro instabil- tion economies (see Table 9). ity rose perceptibly in the majority of middle income countries, i.e. countries that were already more suscep- As for the changes in incidence of income or con- tible to greater volatility. With the possible exception sumption transfers the evidence is inconclusive, with of South Asia, globalisation appears in fact to have in- favourable examples in some cases and negative ones creased the instability of GDP/c (Figure 3), private con- in others (as in the case of Viet Nam, see Table 9). sumption (Prasad et al. 2003), investment, FDI/GDP,

Pathways of transmission and evidence of impact 37 Figure 3: Number of financial crises, 1975-2002

0 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

Source: Caprio and Klingebiel (2003). Note: the chart would most likely have a different appearance if the intensity of the financial crises was plotted rather than the number. However, there is no readily available information on this.

Table 10: Average standard deviation of GDP / c growth rate by country groups 1960-2005

1960-1970 1970-1981 1982-1990 1990-2005 Low income 4.69 6.32 4.95 4.58 Middle income 2.77 3.48 4.44 5.62 High income 1.93 2.69 1.91 2.58

Source: authors’ calculation based on WDI (2006). Note: the above values were obtained by computing the decade standard deviation of each country, which were then averaged for the three main areas.

the share of trade taxes in total revenue, government the standard deviation of the GDP/capita growth rate expenditure and the price of major commodities (Big- (measured by the average quadratic deviation from its geri 2004). UNCTAD (2003, cited in Biggeri 2004) trend) is closely and negatively correlated to the survival confirms that the scale of fluctuations in the real prices rate of children of less than five years of age over both of all commodity groups appears to have increased since the short and long term. For instance, if instability rises the 1970s, albeit with considerable variation from one by 5 points, average U5MR rises from 110 to 128 for commodity group to the other.7 the group as a whole, and from 160 to 184 per thousand for the lowest quartile of the country sample. The negative health effects of growing instability have been summarised in a recent paper by Guillaumont et While the traditional sources of instability contrib- al. (2006), who found that in 68 developing countries uted to this deterioration over the last 20 years, the

7 During the 45 months from February 1998 to September 2001, coffee prices fell by 68.5 per cent, while between January 1998 to November 2001 cotton prices fell by 47 per cent.

38 Globalization and Health Knowledge Network Table 11: Number of banking and financial crises

1970-1995 1970-1979 1980-1995 Average Average Average Type of Crises Total per year Total per year Total Per year Balance-of-payments 76 2.92 26 2.6 50 3.13 Twins 19 0.73 1 0.1 18 1.13 Single 57 2.19 25 2.5 32 2 Banking 26 1 3 0.3 23 1.44

Source: Kaminsky and Reinhart (1998) quoted in Biggeri (2004)

main factor was an epidemic of banking, financial out of school, entered hazardous jobs or prostitution and currency crises (Figure 3), which followed the rings, or suffered permanent brain damage in cases of domestic financial deregulation reforms of the mid- acute malnutrition. late 1980s and the capital account liberalisation of the 1990s. Such measures were often introduced in Trade liberalisation has also been a source of macro the absence of adequate information on exposure and micro instability, especially in countries export- levels, prudential regulation, and sufficient regula- ing primary commodities or subject to rapid trade- tory capacity of national and international monetary related restructuring. The evidence on micro insta- authorities. In addition, in a world where countries bility (due to loss of employment, firm failure, and are closely interlinked through flows of trade, remit- indebtedness) is limited but telling. An example of Table 10: Average standard deviation of tances, technology and finance, the instability- af the negative health impact of premature trade lib- fecting one country can quickly spread to the rest of eralisation is offered by the sharp rise in the suicide GDP / c growth rate by country groups 1960-2005 the world. Cross-country contagion is thus far more rate among cotton farmers in the backward Waran- rapid, and so is cross-sectoral contagion, as seen for gal District in Andra Pradesh (Sudhakumari 2003, 1960-1970 1970-1981 1982-1990 1990-2005 instance by the fact that banking crises have in many quoted in Cornia 2004). In July 1991 the Indian Low income 4.69 6.32 4.95 4.58 cases triggered ‘twin crises’ in the balance of pay- government cut sharply state support to seeds, fertil- Middle income 2.77 3.48 4.44 5.62 ments and currency market, due to capital flights izers, rural credit and output prices, and encouraged High income 1.93 2.69 1.91 2.58 and other effects (Table 11). subsistence farmers to enter the production for export of input-intensive commodities such as cotton. Financial crises dampen economic growth, raise pov- Had markets (including for credit and insurance) Source: authors’ calculation based on WDI (2006). Note: the above values were obtained by computing the decade standard deviation of each country, which were then averaged for the three main areas. erty and therefore affect health status not only in the been complete and efficient, such reforms might have immediate but also over the medium term. Stiglitz lead to a better allocation of farm resources, swift ad- (1998) has shown that countries which were hit by justments to changing market opportunities, faster banking crises over the 1975 94 period grew on aver- growth and better health for rural dwellers. How- age by 1.3 percentage points less rapidly over the sub- ever, following the introduction of these reforms sequent five years compared to countries that did not. the Warangal district experienced a surge in suicides Meanwhile the World Bank (2000) estimated that among small cotton farmers due to rising volatility after the financial crises of Argentina (1987-90), Mex- in world market prices, the absence of any domes- ico (1994-6) and Russia (1996-8) poverty incidence tic insurance mechanisms, declining state support to stabilized at rates of 7-10 points higher than prior to rural activities, and the inefficiency, interlocked-ness the crises. Micro studies on the social impact of fi- and feudal nature of local input and output markets nancial crises are less common but point in the same dominated by moneylenders, which led to consider- direction. For instance, analyses of the Mexican and able rises in input costs. Squeezed between volatile Thai financial crises show that even after the economy world prices and rising costs and indebtedness, sev- recovered, health status was still affected negatively as eral increasingly desperate cotton farmers took their during these acute recessions children had been taken life.

Pathways of transmission and evidence of impact 39 Figure 4: Changes in male life expectancy at birth in relation to a stress index summarizing changes in unemployment, labour turnover and family completeness, Russia, 1989-1993

1 2 1 3 1 4 1 5 1 6 1 7 1 8 1 9 2 0 2 1 2 2 2 3 -2

-3 Central Black Soil Caucasus Volga -4 Volga-Vyatsk

-5 Central Δ W.Siberia Urals Kaliningrad Far-East -6 E.Siberia L E B -7 North-West North

-8 Stress index (Unem ploym ent, Labour T urnover and C hange in the % of M arried Adults)

Source: Cornia (1996)

The most shocking case of rising stress-related mortal- economy, a surge in distress migration related to the ity linked to unexpected and unmitigated macro and search of new jobs, and a rise in family breakdowns. All micro instability is that observed during the difficult these factors interacted with each other to cause a high transition to the market economy of the former Soviet level of stress and a considerable fall in life expectancy Union and Southern Eastern Europe. The health impact (Figure 4). of these developments was dramatic and, for the region as a whole, resulted in an excess mortality of about 10 Lack of institutions and safety nets, preference for million people in the 1990s. The main factor in this rapid liberalisation of domestic and external markets, crisis was the acute psychosocial stress which a large insider-privatization, and the collapse of the state were segment of the population underwent when trying to the main causes of this rise in mortality. The cases of adjust to loss of employment, rapid labor turnover, un- Slovenia, Slovakia, and the Czech Republic represent stable employment in unregulated labor markets, a rise an interesting contrast to that of the former Soviet in ‘undeserved income inequality’, family breakdown Union. In fact, in these countries - where safety nets and distress migration to seek work or political asylum and regulatory institutions were sufficiently developed or to flee conflicts (Cornia and Paniccià, 2000). These and the approaches to domestic and external liberali- various forms of personal instability and the inability to sation and to privatization were gradual and equitable adjust to new circumstances reinforced each other, and - mortality fell sharply in the wake of the transition. interacted negatively with greater alcohol consumption and reduced access to health care. For instance, the fast- 5.4 Income inequality er than average rise in unemployment recorded between 1992 and 1994 in the northern part of Russia caused a There is growing agreement that, after a period of de- high labour turnover, the spread of an unregulated grey cline in the 1950s and 1960s, within-country income

40 Globalization and Health Knowledge Network Table 12: Trends1 in the Gini coefficients of the distribution of income2 from the 1950s to the 1990s for 85 developed, developing and transitional economies

Inequality trend # of OECD # of # of Total # % of % % countries transition developing of world countries populations GDP-PPP countries countries countries Rising 13 24 23 60 71 78 73 Constant 1 1 14 16 18 17 16 Declining 6 0 3 9 11 5 11 Total 20 25 40 85 100 100 100

Source: Cornia and Rosignoli (2007). Notes: Trends in Gini coefficients were interpolated through linear, quadratic and hyperbolic functions. The best results were chosen on the basis of the combination of the best “F”, ‘t’ and ‘R2’ statistics. (2) “per capita household disposable income” in 54 cases, “per capita consumption expenditure” in 9; “gross earnings” in 14. inequality increased – in some cases substantially – over sible to reconstruct an inequality trend between 1980 the last 20-25 years in most countries. Indeed, starting and the early 2000s (Table 12). Inequality fell only from the mid 1970s, and increasingly so since the early in nine small and medium size countries (including 1980s, a reversal in inequality trends was observed in West Germany till 1997, France and Malaysia), and the OECD (Smeeding 2002), Latin America (Szekely it remained broadly constant only in 17. The evidence and Hilgert 1999), the former members of the USSR also shows that in one sixth of the 60 countries that and South Eastern Europe and, to a more limited ex- recorded a rise in inequality, this stabilized once the tent, those of Central Europe (Milanovic 1998). Mean- new liberal policies had been fully implemented. This while, in China inequality rose slowly over 1978-1984, was the case, for instance for the UK (Figure 5), Brazil, but much more rapidly after 1990. A reversal of the Russia, Romania, Moldova and several other European inequality trend was also observed in the economies economies in transition, where due to marked changes of the East Asian miracle, which is famous for having in economic policies income, inequality shifted from achieved equitable rapid export-led growth. And in In- the lower level associated with Keynesian or Socialist dia, the liberalisation of 1991 led to a moderate rise in policy regimes to a higher level typical of liberalized urban and rural inequality and a larger rise in nation- and open economies. Once this policy shift had been wide inequality due to a widening urban-rural income completed, and no other policy changes were intro- gap (Deaton and Drèze 2002, Sen and Himansu 2006). duced, inequality remained constant. Inequality rose less markedly in Sub-Saharan Africa and even less so in the MENA region, where, however, data Such widespread rises in inequality affected the speed scarcity limits the scope of the analysis. of growth (section 3) as well as health status. Initial regression analysis on the 12 Russian macro regions for The latest global analyses of within country inequal- 1989-1994 (Cornia, 1996) suggests that a 20-point ity point to an overall increase. Jantti and Sandstrom increase in the Gini coefficient was associated with a (2005) analyse the latest version of the World Income reduction in life expectancy at birth of 1.5 years, while Inequality Database, which contains data for 115 a 10 per cent rise in unemployment was associated countries, and found that in most of them inequality with a reduction of 3.5. had risen mainly because of an increase in the income share of the richest fifth. In turn, Cornia and Rosignoli There is considerable debate on the causes of this increase (2007), using the same dataset, suggest that income in inequality. Some of the hypotheses focus on South- inequality rose – albeit to different extents, with differ- North trade, migration, and technological change. These ent timing and most probably with different effects on explanations apply in some cases, but cannot be genera- health - in 60 of the 85 countries for which it was pos- lised to explain the rise in inequality observed in very

Pathways of transmission and evidence of impact 41 Figure 5: Trends in the Gini coefficients of income inequality in China and household disposable income in Great Britain, 1960-2003

Key: a= gross income For China: For Great Britain: b= income a= ndj;ing means ‘scale unadjusted gross income’ d= hsh;ind means ‘scale adjusted disposable income’ e= gross income b= hsh;in means ‘scale adjusted household income’ (net income after taxes and transfers). d= net disposable income e= mss;ing means ‘scale undefined gross income’

42 Globalization and Health Knowledge Network Figure 6: Changes in inequality and suicide rate in transition countries, 1989-97 60

Russian Federation

Moldova

Latvia 40

Poland Slovenia Lithuania Romania Bulgaria Estonia 30

Kashakstan

Belarus 20 Slovakia Czech Republic East Germany % change income inequality 1989 - 1997 10 Hungary r = 0 .6 3 ; p < .0 0 1

0 -40 -20 0 20 40 60 80 % change suicide rate 1989-1997 Source: De Vogli (2004)

different country settings (Cornia 2004). Policy-driven rapidly, and in the European economies in transition globalisation may therefore be the main driving factor. unemployment rose by 10 million units between 1989 This point is discussed further in Section 8. and 1996 alone, while highly unstable and often hazardous employment in the informal sector skyrocketed As already noted, whatever the reason for its rise, in- (Tchernina 2000). In contrast, most of East Asia ex- come inequality can affect health status via the acute perienced a sustained growth of formal sector employ- stress pathway. Figure 6 shows that in the economies ment generating – ceteris paribus – a positive health ef- in transition, a sharp increase in inequality following fect (van der Hoeven and Saget 2004). the adoption of botched liberalisation policies led to a rise in the suicide rate. Formal sector employment fell as import liberalisation and privatisation reduced employment in the 5.5 Trends in formal and informal protected and state sector, while slow GDP growth employment and in unemployment did not generate enough new formal sector jobs to ac- commodate retrenched workers and a rapidly rising Employment trends differ substantially across regions. labour supply, despite devaluation, currency convert- In five Southern African countries, the percentage of ibility and lower deficits (van den Hoeven and Saget formal sector employment fell between 1990 and 1995 2004). Indeed, GDP growth has slackened over the from 17-29 percent to 13-25 percent, while in predom- last two decades partly because of the contractionary inantly middle income Latin America it fell from 49 to stabilisation programs of the 1980s and 1990s which 43 percent (van der Hoeven and Saget 2004). Mean- stymied rather than stimulated growth and employ- while in South Asia informal employment expanded ment creation.8

8 The IMF and World Bank have argued in the past that without the structural adjustment programmes they asked the countries to introduce as part of their conditionality, things would have been even worse. While there is no doubt that countries facing macroeconomic crises have to adjust, it is by now abundantly clear that the Polak model, which still guides the adjustment approach of the IFIs, suffers from several analytical shortcomings (short term horizon, overkill, over-reliance on restrictions rather than on stimulation, etc.) some of which have been corrected in the more recent adjustment program of the IFIs. Moreover countries where there have been more successful adjustments, have often followed different approaches (for more details see Stiglitz 1998 and Cornia 2006).

Pathways of transmission and evidence of impact 43 Finally, in some countries, informal sector employ- shorter period following the financial crisis of 1997-8. ment at low wages rose due to loss of competitiveness, The unemployment rate rose from 2 percent prior to the as – in the wake of the liberalisation of the capital ac- crisis, to 8.4 in the first three months of 1999, to return count and the ensuing appreciation of the exchange to 4.6 by 2001 (Cornia 2004). In addition, the share rate - domestic firms either delocalized or subcontract- of part-time and daily workers not covered by social ed productive tasks to informal sector firms paying insurance jumped from 42.5 to 52.5 percent between lower wages. Overall, with the liberalisation of labour l996 and 2000 (KLSI 2001). As a result, between 1996 markets, the trend has been towards greater employ- and 1999, the number of work-related deaths rose by ment flexibility, mobility and job instability, but in 10 percent, occupational ailments by 18, and ischaemic many cases this approach has raised unskilled employ- diseases by 100 percent between 1996 and 2001, while ment at low wages. Obviously analyses of changes in suicides rose from 5 to 9 per 100.000. employment and unemployment make sense only for middle and high income countries of the OECD, in The biggest increase in death rates following changes transition and from Latin America. In most low in- in the labour market (and other spheres of life) was come developing and transitional economies, under- observed in the 1990s in the economies in transition employment is the main problem. of the former USSR. In Russia, Ukraine and Belarus the standardized death rate doubled for males in the What was the health impact of the above changes in 20-40 age groups. Empirical evidence suggests that labour market and macroeconomic policies and condi- one of the main triggers of this health crisis was the tions? The macro health impact of growing informalisa- sharp rise in unexpected unemployment, job insecuri- tion and job-instability is difficult to measure, but there ty and income inequality that followed rapid econom- is micro evidence that loss of employment, informalisa- ic liberalisation. Uncertainty, depression and mental tion and underemployment affected the health status disorders, alcoholism, domestic violence, deaths due of certain groups of the population. For instance, the to cardiovascular and violent causes and suicides rose greater labour market flexibility introduced in the early among an underclass of middle-aged male workers, 1990s to revitalize the Japanese economy appears to collective farmers and party cadres with limited edu- have affected the health status of the groups affected. cation and skills, who lost their jobs or were obliged Companies were allowed to scrap the old system of to enter unstable and low paid jobs, or migrate un- lifetime employment in favour of a more flexible one der stressful conditions (Cornia and Paniccià, 2000). rewarding productive workers. However, as a result of A confirmation of the negative effect of unemploy- this change, unemployment – unknown for decades – ment on health is provided by the Russian financial rose steadily from close to zero in the late 1980s to 5.4 crisis of August 1998, which triggered a 100 percent percent in 2002, while rapid job turnover became more devaluation of the rouble, a huge price increase, a sud- common. The health impact was severe, as workers den three-point rise in unemployment, which led to were not geared to the the possibility of losing their job, a three year decrease in male life expectancy between a particularly shameful event in a culture underscor- 1997 and 2000. ing lifelong attachment to a company. While mortality rates for most causes – including for heart diseases – de- 5.6 Inflation and prices of basic goods clined slowly or remained broadly unaltered, the number of suicides grew by almost 80 percent between 1990 Broadly speaking inflation declined over most of the and 2000. For the decade as a whole there were some 1980-2005 period (Figure 7), partly because of the 80,000 excess suicides among the new unemployed, as emphasis placed by the IMF on achieving low infla- well as the entrepreneurs and managers of bankrupted tion. companies who were unable to adjust to the shame, de- jection and stigma felt by those who had lost their job The liberal policies implemented over the last quar- and their role in society (Lamar 2000). ter century also emphasize measures affecting the price of goods (food and medicine) which are very The case of South Korea has several similarities with important for health. Firstly state subsidies to key that of Japan, but in this case health deteriorated over a consumer goods were withdrawn as part of price lib-

44 Globalization and Health Knowledge Network Figure 7: Inflation spline (i.e. the median inflation of countries examined) for the period 1960-2002

.15

Median spline.1

.05

19 60 1 970 19 80 19 90 20 00 ye ar

Figure 8: Trend in the ‘food-price/consumer price’ index by country groups 1980-2005

HIGH LOW 1.2

.6 Median spline MIDDLE TRANS 1.2

1980 1990 2000 20101980 1990 2000 2010 ye ar Graphs by region2

eralisation. In addition, food prices rose faster than 5.7 Taxation, public expenditure on health the CPI in countries which devalued their nominal care, and approaches to its financing exchange rate. These measures should have led to a situation where food prices rose faster than the aver- In the industrialized countries, health expenditure age inflation rate, but examination of the CPI and rose because of the high income elasticity of health food price data included in the GHND does not re- care demand and population ageing. It is difficult to veal a general trend to support this (Figure 8). While generalise about the situation in developing countries. the relative price of food appears to have risen in the Stringent stabilization policies led to declining bud- transition economies, it generally followed a down- get deficits, as the spending capacity of governments ward trend. became increasingly determined by the tax revenue

Pathways of transmission and evidence of impact 45 Figure 9: Average un-weighted tariff rates (percent) by region, 1980-1998

Source: World Commission on the Social Dimension of Globalisation 2004, World Bank, Global Economic Perspectives (2001) (background presentation).

Table 13: Trends in Central Government Revenue(CGR)/GDP ratio

Region Average unweighted Central Government Average Unweighted change Number of countries experiencing Revenue as a share of GDP (CGR/GDP) in CGR/GDP) 1990-2005 changes in CGR/GDP 1990 1995 2000 2005 Increase Decrease No change East Asia (6) 14.0 14.4 13.4 …. - 1.3 2 4 0 Latin America (12) 13.7 13.4 13.4 …. 0.0 8 4 0 MENA (6) 15.5 15.8 16.8 …. +1.2 3 2 1 South Asia (4) 12.8 13.1 11.3 …. - 0.9 1 3 0 SSaharanAfrica (8) 12.8 13.1 11.3 …. 0.0 2 6 0 Transition Europe(9) 15.3 16.9 16.6 …. +1.5 5 3 1 Total (43) …. 21 22 2

Source: Authors’ elaboration based on WDI (2006)

they collected. The reduction of trade tariffs following tion led to the informalization of the economy which, trade liberalization (Figure 9) also affected the spend- by itself, makes revenue collection more difficult. Yet, ing capacity of liberalizing governments, as the drop some countries were able to compensate for decline in revenue from trade taxes was not always fully com- in revenue by introducing VAT or increasing royal- pensated by revenue from other taxes. ties. The net impact of the changes in tax reforms varied from country to country (see the data in Table 13 Furthermore, in the new world of mobile capital and showing trends in central government revenue). immobile labour, developing countries wishing to at- tract FDI introduced measures such as reducing the While trade taxes fell everywhere, the overall effect rates and progressivity of income tax, providing tax of globalisation on the ratio of revenue to GDP was breaks and granting subsidies. In addition, globalisa- mixed. Central government revenue as a share of

46 Globalization and Health Knowledge Network GDP appears to have fallen in the majority of the East 5.49 percent (Viet Nam) of total household income, Asian, South Asian and Sub-Saharan countries but and that inclusion of such unaccounted costs in the to have risen slightly in two thirds of Latin America poverty line would increase the number of the poor by countries (possibly following the return to democracy 78 million people, and the average poverty rate by 2.7 and consequent growth in pressure for redistribution) percentage points. The paper also cites the examples and in the economies in transition where new tax sys- of Sri Lanka, Malaysia and Indonesia, where govern- tems were introduced as part of institutional reforms. ments were able to raise sufficient revenue through taxation to keep charges for public health care to a The analysis of the trend in tax/GDP ratios in devel- minimum, or to introduce targeted exemption by is- oping countries carried out by Chu et al.(2000) points suing health cards which substantially reduced the im- to an average drop of one percentage point in the tax/ pact of user charges and other out of pocket expendi- GDP ratio over the 1980s-1990s period, compared to tures on poverty rates. a rise by 1.6 points between the 1970s and 1980s. As a result, public health expenditure declined in several Similar or even more striking results have been re- countries (e.g.China between 1978 and 1990, and In- ported for the transition economies of the former So- dia after 1991), although in other cases (the Middle viet Union. Even in the highly successful economies East and North Africa) it did not change much or even of China and Viet Nam, the health financing reforms increased. Jha (2006) also finds a negative trend in tax led to a decrease in access to health care by the poor. collection relative to GDP, after controlling for GDP/c With the Doi Moi market reforms of 1986, Viet Nam through a panel fixed effects regression of fifteen large reformed the financing and provision of health care developing countries over the period 1985–98. from a state-financed system based on community health centres and hospitals, to one based on health Thirdly, access to health care has also been influenced insurance for those working in the formal sector, user by the health sector financing reforms adopted during fees in public hospitals, legalisation of private medi- the last 15 years. These reforms were often introduced cal practice and liberalisation of drugs sales. These in the wake of the fiscal crises of the first half of the reforms led to a drop in public health expenditure, 1980s and of the structural adjustment programs pro- and a rise in private health costs, as the share of the moted by the IFIs to tackle them. Such reforms pro- latter in overall health expenditure rose from 59.6 to moted a market-based concept of health sector reform 74.2 percent between 1995 and 2000 alone, while and aimed at limiting the role of the state in health public expenditure on health fell from 1.57 to 1.34 of provision, increasing the efficiency of health expen- GDP over the same period. Microeconomic analyses diture, introducing user fees in public health institu- based on the LSMS surveys of 1993 and 1998 (Tiberti tions, decentralizing services provision to local govern- 2006) show that these reforms led to a rise from 6 to ments and non-governmental organizations, opening 17.6 percent in the proportion of sick people not seek- up to private providers, privatizing government hospi- ing care, while the proportion of sick people visited at tals or parts of thereof, and introducing health insur- home by a health practitioner fell from 6 percent to ance. Most of these reforms (user fees, privatisation, 2.3 percent despite a rise in the income of most of the health insurance), however, remain highly controver- population. The share of respondents who were hospi- sial due to their effects on access of the poor to health talized also fell, while the number of people relying on care. Research has shown that market-based reforms self-care rose sharply. These results were particularly in health sectors of sub-Saharan Africa over the past pronounced for children of 0-5 years of age, and in 20 years have reduced service utilization rates (Reddy the urban areas. and Vandemoortele, 1996). A similar review by van Doorslaer et al. (2006) based on household budget The Chinese evidence is even more worrying. In the data for 11 Asian countries representing 48 percent socialist era 3% of GDP was spent on a universal of the world’s population, shows that out-of-pocket health care system. With the introduction of reforms payments (medical fees, user charges for public care, in the 1980s and their acceleration in the 1990s, this purchases of medicines, insurance co-payments and so system was replaced by one which fostered the privati- on) absorbed on average between 1.37 (Malaysia) and zation and marketization of health services. The share

Pathways of transmission and evidence of impact 47 Table 14: Access to health care and financial difficulties in China, 1993 and 1998

Year Big Cities Medium Cities Small Cities Rural areas I II III IV % of patients not seeking medical treatment due to financial difficulties 1993 3.2 2.4 9.6 15.1 21.4 19.5 24.4 1998 36.7 23.5 42.9 30.1 31.7 42.3 38.7 % of patients not hospitalized when they should be due to financial difficulties 1993 34.1 33.9 53.4 47.9 63.5 61.1 67.7 1998 53.1 58.4 70.8 63.8 54.1 70.2 69.4

Source: Lu (2006) based on data from the 2nd National Health Service Survey of 1998. Notes: I-IV refers to groups of rural areas ranked according to income per capita.

of GDP spent on public health expenditure dropped 5.8 Migration and family arrangements to less than one percent, and responsibility for public services was transferred to local governments, leading In the CIS countries alone, nine million people, wheth- to a deterioration of health services in the poorer ar- er seeking work or political asylum, fleeing conflicts or eas. After almost 30 years of transition, the population returning from abroad, have migrated since 1989. For covered by government or other forms of health insur- many of them, migration entailed considerable material ance is estimated at half of the urban workforce, while hardship, disorientation and loss of control, breakdown only 10 percent of the rural population is covered. For of social relationships, redefinition of survival strategies, most people, out of pocket payments have become the and housing problems. Distress migration has also been only way to secure medical care. The share of indi- common in many other countries, and the stock of mi- vidual cash payments in total health expenditure, rose grants worldwide has risen to 190 million. from 20.4% in 1978 to 55.9% in 2003, while that of government budget and various public health insur- While migrants (particularly distress migrants) may ance schemes fell from 32.2 to 17.0 percent and from suffer a decline in their life expectancy, migration may 47.4 to 27.2 percent respectively (Lu 2006). also help improve the health status of the population of the countries of destination. This effect – seldom As a result, - despite a rise in total health expendi- discussed in the literature - may be due to better staff- ture to 5.6% of GDP – health care coverage in China ing of health services, better care of children and the has been dramatically reduced. Surveys carried out in aged in middle and high income destination countries, 1988, 1995 and 2002, suggest that the share of families especially in the OECD and Gulf states. For example, with members who are not covered by health care rose 30 percent of the Ghanean, 20 percent of the Ugan- from 17.1 to 62.7 percent between 1988 and 2002. dan, 15 percent of the Ethiopian, 43 percent of the Although commercial and other types of insurance Liberian trained physicians work in the US or Canada schemes have been created, their coverage is limited alone (Bach 2006). Likewise, over half of the doctors and they are unable to absorb those who have been registered in the UK in 2002 were trained outside the squeezed out from the public health care systems, and UK, and especially outside Europe (ibid.) an increasing proportion of the population has been left without any coverage (Table 14). The effects of Secondly, migrants may work in low-skilled jobs which these measures on mortality are negative, but at the the local population no longer fill, contributing in this same time health outcomes were positively influenced way to overall economic growth. Thirdly, migration in- by the rapid growth in income. creases the supply of labour and in this way acts to

48 Globalization and Health Knowledge Network Table 15: Cardiovascular interventions* in OECD countries (DDD/1000 inhabitants/day)

1980 1985 1990 1995 2000 2005 Australia …. …. 416.9 439.3 387.6 479.4 Denmark …. 167.7 161.6 192.6 250.2 361.8 Germany …. 274.7 317.8 332.6 373.2 441.9 Greece …. …. 140.3 194.0 311.8 455.5 Sweden 197.9 200.4 206.5 225.7 289.7 397.5

Source: OECD Health database (2006). Notes: *cardiac glycosides, antiarrythmics, anti-hypertensives, diuretics, beta and calcium channel blockers, agents acting on the renin-angiotensin system, cholesterol and triglyceride reducers.

Figure 10: DPT3 percentage immunization rate, 1980-1999

Source: authors’ elaboration on UNICEF data

contain the growth of wages in the countries of destina- 5.9 Technical progress in health tion, thus enhancing competitiveness in the traditional sector.9 On the other hand, the short and long term How has globalisation – both endogenous and poli- health impacts of migration and migrant remittances cy driven – affected technical progress in health? To in the countries of origin are more ambiguous. While answer this question, three sub-questions need to be migration of skilled personnel (especially of doctors, addressed, i.e. (i) has liberalisation-globalisation im- nurses and allied health professionals) contributes to proved incentives to produce new medical discover- the ‘brain drain’, remittances may allow improved con- ies to solve the health problems of the advanced and sumption of food and drugs, thus contributing to better developing countries? (ii) have international trade and short term health outcomes. The long term growth ef- technology transfer policies facilitated the transfer of fect of remittances is, however, much less clear. old and new knowledge and drugs from the more

9 Containment of wages in this sector may, however, raise wage and income inequality.

Pathways of transmission and evidence of impact 49 Table 16: Prevention of mother-to-child transmission with antiretroviral prophylaxis

UNGASS Country Report 2003, 2005 Coverage Survey Percentage of HIV-infected pregnant percentage of HIV-positive pregnant women receiving a complete course of receiving antiretroviral prophylaxis antiretroviral prophylaxis to reduce the risk of mother-to-child transmission 2003 2005 2003 2005 Botswana 34.0 18.8 Burundi 1.2 13.2 2.4 Cote d’Ivoire 1.5 4.4 4.3 Kenya 1.0 3.4 9.3 27.0 Lesotho 5.3 5.1 Namibia 7.0 1.3 25.0 17.4 South Africa 1.0 8.9 78.7 14.6 Uganda 4.6 6.6 12.0 25.9 Zambia 6.3 25.0 4.0 Zimbabwe 4.1 6.6 4.4

Source: UNAIDS Report 2006

to the less advanced countries? (iii) has domestic lib- by reducing tariffs and trade costs. Likewise, a cheaper eralisation improved access to the newly transferred and faster transmission of information via international technologies? television networks and the internet sped up the transmission of information on good health practices. As a Regarding the first question, it is difficult to prove result, there is evidence that mortality declined in the empirically that the liberal regime of the last twenty advanced countries, thanks mainly to the diffusion of years has led to an acceleration in health discoveries new knowledge about the risks of cigarette smoking, compared to the prior two decades, although this has and the spread of new procedures and drugs which re- most likely been the case in the advanced countries. duce mortality due to cardiovascular diseases (Deaton What is abundantly clear, in contrast, is that health 2004). Table 15 illustrates this point by showing the research continued to focus on health conditions al- steady increase in the coverage of procedures and drugs most exclusively found in the industrialized societies. that reduce cardiovascular mortality. As noted in Labonté and Schrecker (2006), only 10 percent of health research expenditure is allocated to The last twenty years have also seen a diffusion of old diseases that account for 90 percent of the global bur- and new drugs and medical technologies in develop- den of diseases, and which are most common in the ing countries, including vaccines, oral rehydration developing nations. As a result, of 1393 new drugs salts (ORS), nutritional supplements (iron, folate, vi- patented between 1975 and 1999, only 16 were for tamin A and B12), antibiotics, aspirin and other drugs tropical diseases and tuberculosis; and as of today, no for which the patents had expired. The spread of these vaccine has been discovered against malaria, a disease low-cost health technologies played a key role in re- that accounts for 10 percent of all deaths in sub-Saha- ducing infant, child and overall mortality. As shown ran Africa (see Table 4 above). in Figure 10, the coverage of DPT3 immunisation rose sharply in all regions between 1980 and 1990, al- Second, external liberalisation had a positive effect on though immunisation rates stagnated after 1990 in all the transmission of health knowledge and technology regions, and declined in sub-Saharan Africa because

50 Globalization and Health Knowledge Network of the fall in international support for these programs, developing countries, meant that it has not been pos- and the reluctance or inability of governments to fi- sible to control the mortality due to AIDS, TBC and nance their costs. malaria and to reduce the 10.5 million child deaths recorded each year due to readily preventable diseases There was less transfer of the more sophisticated drugs such as measles, diarrhoea, and pneumonia. Different and technologies, including beta-blockers, screening intellectual property regimes, aid and health financing equipment, cardiac units and, until recently, antiret- policies – part of a different globalisation approach roviral drugs. Indeed, while lower trade tariffs reduced – are needed to ensure that the advances of modern the costs of imported drugs, the adoption of a more health technology can eventually reach all the poor. stringent intellectual property regime embodied in the TRIPS agreement (which guarantees a longer period 5.10 Fertility rates and dependency ratios. of monopolistic regime, more products/processes and higher royalties and fees), raised the cost of newly pat- Fertility rates have declined massively in all develop- ented foreign drugs and medical technologies, and there- ing regions over the last quarter century, with the ex- fore excluded – more than before – people from the use ception of sub-Saharan Africa and a few countries in of old and, especially, newly patented drugs. As argued South and West Asia. As a result, dependency ratios by Deaton (2004, p. 30) ‘…There is clearly a long way improved in many countries. Over 1980-2005 the to- to go before the [health] habits and technology of the tal fertility rate (TFR) fell between 33 and 51 percent rich countries are fully adopted even in middle-income in East Asia, Latin America, South and East Asia and countries’. A concrete example of this is provided by the MENA (Table 17), with much faster declines in coun- limited coverage of nevirapine (a comparatively low-cost tries such as Bangladesh and Morocco. This contrib- antiretroviral that reduces the risk of mother-to-child uted to a rapid reduction in the number of high-risk, transmission of the HIV virus). Although the drug was high parity, and closely-spaced births, and of births to discovered in 2000-1, its coverage rate in countries with young and old mothers, all of which contributed to a high HIV adult prevalence remains very low, and in the decline in IMR and U5MR – and to a lesser extent some cases is declining (Table 16). to an improvement in LEB – in several developing countries. In contrast, changes in TFR and dependen- Thirdly, the adoption of newly transferred health cy ratios in the transition and advanced countries were technologies and information is not automatic: it more limited, but may have influenced the changes requires changes in health behaviour, an adequate in health status in the former. In turn, in the OECD institutional set up to facilitate their diffusion, and countries, a worsening of the old age dependency ratio access to finance. As noted in sub-section 4.7, the entailed a large increase in health expenditures that liberal philosophy which may have created greater was not accompanied however by a surge in old age incentives for health research also favoured in most mortality rates. countries an approach to the financing of health care based on user fees and other cost recovery measures 5.11 Smoking, drinking and obesity in public health facilities, health insurance, and privatisation of health, i.e. measures that increasingly Has globalisation reduced or increased the risk factors placed the burden of health financing on the house- for health associated with smoking, drinking and obe- hold. As a result, out-of pocket health expenditure sity? As noted above, these are particularly prevalent accounts for a large and growing share of overall in the developed countries, although they also affect a health expenditure. growing share of the affluent class and even the poor in low- and middle-income countries. Indeed, some In conclusion, the large potential benefits of the of the less favourable changes have occurred in the North-South transfer of health technology and knowl- middle-income countries. edge were not fully realized due to policy changes in the field of intellectual property rights and health The International Mortality and Smoking Statistics financing, especially in the 1990s. These policies, as data base contains survey-based estimates of age- well as weak aid policies and domestic institutions in adjusted prevalence rates of smoking for five year

Pathways of transmission and evidence of impact 51 Table 17: Average population-weighted Total Fertility Rates by region, 1960-2005

Year East Asia OECD LAC MENA S&SEA SSA Tr. Econ Russia 1960 6.00 2.86 6.09 6.83 6.57 6.73 2.44 2.56 1965 5.52 (-8) 2.68 (-6) 5.67 (-7) 6.51 (-5) 6.15 (-6) 6.73 (0) 2.35 (-4) 2.54 (-1) 1970 5.39(-10) 2.34 (-13) 5.26 (-7) 6.12 (-6) 5.82 (-5) 6.70 (0) 2.34 (0) 2.50 (-2) 1975 3.78*(-70) 1.92 (-18) 4.65 (-12) 5.73 (-6) 5.37 (-8) 6.69 (0) 2.31 (-1) 2.48 (-1) 1980 2.96(-22) 1.79 (-7) 4.15 (-11) 5.53 (-3) 5.12 (-5) 6.66 (0) 2.23 (-3) 2.46 (-1) 1985 2.69 (-9) 1.72 (-4) 3.66 (-12) 4.91(-12) 4.57 (-11) 6.47 (-3) 2.15 (-4) 2.35 (-4) 1990 2.39(-11) 1.71 (-1) 3.25 (-11) 4.09 (-17) 3.95(-14) 6.14 (-5) 1.93 (-10) 2.24 (-4) 1995 2.11(-12) 1.67 (-3) 2.94 (-10) 3.29 (-20) 3.58 (-9) 5.73 (-7) 1.51(-22) 1.73(-23) 2000 2.01 (-5) 1.65 (-1) 2.71 (-8) 2.88(-12) 3.20(-11) 5.38 (-6) 1.29 (-15) 1.53(-12) 2005 1.97 (-2) 1.65 (0) 2.59 (-4) 2.76 (-4) 3.09 (-3) 5.25 (-2 ) 1.25 (-3) 1.51 (-1)

Source: authors’ estimates based on WDI 2006 (World Bank). Total fertility rate is the number of children that would be born to a woman if she were to live to the end of her childbearing years and bear children in accordance with prevailing age-specific fertility rates in each period. Note: * China’s TFR fell from 5.5 to 3.4 between 1970 and 1975. Note: the numbers in parentheses are the percentage changes for the five year periods.

Figure 11: Trends in alcohol consumption by region

15 HIGH LOW

0 Median spline 15 MIDDLE TRANS

1960 1980 2000 1960 1980 2000 ye ar Graphs by region2

periods from 1951–1955. These show that, after rising prevalence continues to rise for both men and ing from 1960 until about 1980, smoking incidence women. These smoking patterns are clearly reflected among men has steadily declined in many OECD in international patterns of mortality from lung-can- countries as a result of awareness campaigns, a va- cer (Deaton 2004). For instance, lung cancer mor- riety of prohibitions, class actions and court rulings tality among women started from much lower levels requiring the tobacco companies to provide compen- than for men, but is still rising in many countries. sation to people contracting smoking-related diseas- As lung-cancer mortality represents approximately es. However, the latest survey data suggest that the one fifth of mortality from cardiovascular diseases, share of women smokers outside the Anglo-Saxon the above changes in smoking prevalence have had a countries is still rising; and in Eastern Europe, smok- large impact on death rates in these countries.

52 Globalization and Health Knowledge Network Figure 12: Number of conflicts in the last decades 1960-200210

Source: authors’ elaboration based on CRED data, www.cred.be

Data on the developing countries are scattered but is rising in developing countries at a much earlier stage the tendency is towards a rise in smoking incidence in of economic development than implied by the ‘diseas- many countries, especially where the tobacco TNCs es of affluence paradigm’. Data on the consumption relocated following liberalisation of FDI and growing of alcohol over the last twenty years show an overall prohibitions in advanced countries, and despite the at- growing trend, apart from in the low income nations tempt to adopt an international covenant on the trade, (Figure 11). marketing and control of tobacco. 5.12 Exogenous developments and random shocks11 Equally worrying has been the impact of the liberalisation of FDI in the processing and distribution of A number of shocks have affected health trends in sev- Western processed foods and drinks rich in salt, sugar eral regions in the last quarter century. In countries and fat and low in fibre (Hawkes 2004). Supermarket with weak institutions for redistribution and conflict purchases of Western processed foods in developing management, the impact of these shocks was often countries already account for a quarter of total food magnified, as they triggered a rise in conflicts which purchases in these countries. As a result of this tran- reduced growth and increased the negative health im- sition from traditional to Western diets, the link be- pact. tween obesity and economic development becomes evident even at comparatively low income levels (Ez- (i) Natural disasters and famines. Major food shortages zati et al 2005). One study, based on a review of a large were frequently recorded mainly in sub-Saharan Af- number of clinical studies from over 100 countries, rica, particularly over the 1983-1985 period and in argues that this nutritional transition has led to a rapid the early 1990s. At present, an important part of the increase in body mass index, obesity, high cholesterol, population of Southern and Eastern Africa depends blood pressure, cardiovascular diseases and diabetes on emergency food aid for its survival. Severe famines particularly among the urban poor. As a result of these were experienced by Sudan and Southern Eastern Af- globalisation-related changes, cardiovascular mortality rica in 1990-1992, in North Korea in the mid 1990s,

10 Contrary to the CRED database drawn on for Figure 12, the 2005 Human Security Report suggests that the number of internal conflicts and humanitarian emergencies has declined since the mid-late 1990s. All sources agree, however, that there was an increase in the number of such events from the mid 1980s to the mid-late 1990s. 11 Most exogenous shocks analysed in this section are due in part to policy factors. Natural disasters could be seen as the result of the present unsustainable development model, conflicts or instability could be attributed to the TNCs operating in the commodity sector, or the abandonment of ‘client states’ by the major superpowers, and HIV-AIDS to economic policies that raise inequality. While there is a grain of truth in all these claims, it is also very hard to convincingly link these events to globalisation policies.

Pathways of transmission and evidence of impact 53 and – in a more moderate way – in several Sahelian the number of external and internal refugees has in- countries in 2005. In some cases, civil unrest prevent- creased continually. In 1995, there were 5.2 million ed the delivery of humanitarian aid, and hunger was refugees in sub-Saharan Africa as a whole, with the particularly severe in war zones. This explains why im- biggest group being in Zaire (1.3 million), followed provements in predicting famines have not resulted in by Tanzania (0.7 million), and Guinea (0.6 million) parallel improvements in its prevention. (World Refugee Survey 1996:4-5). The number of internally displaced people was even greater and was (ii) War and displacement. The last quarter century- wit estimated at over 8 million people in Africa alone. nessed a sharp rise in the number of local wars and complex humanitarian emergencies. These increased (iii) AIDS and other diseases. The precise mortality impact from 25 per year in the early 1980s to almost 60 in of HIV-AIDS is large but difficult to estimately pre- the mid-late 1990s and early 2000s. (Figure 12). The cisely. AIDS currently kills about 2.2 million people a factors behind this rising number of conflicts include year. Some 80 per cent of the 30 million HIV infected mounting polarisation among ethnic, religious or eco- people live in sub-Saharan Africa, and an even higher nomic groups, a weakening of the state, and fights share of the total deaths is estimated to have occurred over natural resources. Meanwhile, the smuggling of there. Unless treated with antiretroviral drugs, the dis- valuable commodities was both a major incentive to ease is always fatal, although its impact on mortality continue local wars and a way to finance them. varies according to the nutritional levels of the populations affected and the availability of drugs to delay Sub-Saharan Africa and the former Soviet Bloc were the onset of AIDS. Already now, baseline mortality the two regions most affected, but problems were also rates for the population groups affected have already common in Afghanistan, East Asia (e.g. Indonesia, doubled or tripled. The incidence is generally higher in Timor) and in Latin America. For instance, in 1994, Eastern than Western Africa. In the late 1990s, urban there were no less than 13 full-fledged wars in SSA, af- HIV prevalence rates among adults in Eastern Africa fecting Angola, Ethiopia (Eritrea and Tigray), Liberia, ranged from 12 per cent in Tanzania to 35 per cent in Mozambique, Rwanda, Somalia, South Africa, Sudan, Rwanda. In West Africa it ranged between 1 per cent Uganda, Burundi, Sierra Leone and Zaire. If a less re- in Cameroon to 15 per cent in Cote d’Ivoire. Rural strictive definition of war is adopted, the number of prevalence rates were estimated to be roughly 20-30 armed conflicts underway during this period rises by per cent of urban rates, with the exception of Zambia about one third. Most of these conflicts are increas- (where mortality rates, including for under-fives, rose ingly due to internal problems (struggle for power, drastically in the 1990s), Zimbabwe, Tanzania and secession, or autonomy), and involve non-state actors, Central African Republic, where rural rates are equal such as armed clans. This aggravates the health impact to two thirds of urban rates. as battles are fought in the midst of the civilian population.

Though the measurement of the mortality impact of civil wars is difficult, existing databases show that the numbers are considerable, i.e. between 200-500,000 thousand casualties in Rwanda and 100,000 in Ango- la in 1994, 100,000 in Burundi in 1993, 100,000 in Mozambique in 1992, about half in Liberia in 1993, and smaller but non negligible numbers in Sudan and South Africa in 1993.

An increase in the number of refugees without adequate access to food, fresh water and health care has also had a negative effect on the health status of the sub-Saharan African population. Since the late 1970s,

54 Globalization and Health Knowledge Network Globalisation and Health: Pathways of Transmission and Evidence of Impact

6. Econometric estimates of LEB, IMR, U5MR models

his section tests econometrically the relation in a limited number of cases, from national sources between health status and its determinants as or the specialised literature. At present, the GHND Tsummarized in Table 5. Before doing so, we describe comprises national level data for 136 countries and the dataset which was constructed in order to carry ten quinquennia (1960, 1965, …, 2005) and, for the out this analysis. European economies in transition, annual observations for 1980-2005. For the latter, quinquennial data 6.1 The GHND dataset for the 1960-1980 period are unavailable, making it necessary to rely on annual data when estimating the Table 18 summarizes the variables which the litera- ‘regional mortality model’ for this group of countries. ture suggests exert a significant influence on mortality via the five pathways discussed in Section 4. Inspired If data were available for all variables and for all the by this framework, we built the Globalisation-Health countries and years set out in Table 5, the ‘global quin- Nexus Database (GHND) that includes variables con- quennial GHND dataset’ would have 1360 (136x10) cerning mortality, its determinants, indexes of the pol- complete data-strings. However, missing data, espe- icies that may influence them (available mainly for the cially for the beginning and end of the period, mean Latin American and Eastern European countries), and that the number of country-years for which data are random shocks affecting health status. Data for these available for all variables amounts to 583. Table 18 variables was compiled for the 1960-2005 period, a pe- provides data on the ‘data frequency’ of each variable riod that for sake of simplicity can be divided into the included in the regression analysis. When missing data Second Golden Age of Capitalism (1960-80) and the concerning a ‘qualitatively well-known’ phenomenon Second Globalisation Era (1985-2005). The data in threatened to reduce the number of complete data- the GHND are taken from a variety of databases and, strings, data were filled in by interpolation or on the

Pathways of transmission and evidence of impact 55 Table 18: Summary of main variables included in GHND and used in regression analysis

Variable type Variable Name Variable Description Main source Data Density Mortality U5MR Under five mortality rate (per 1000 life births). WDI 94% (29%r) IMR Infant mortality rate (per 1000 life births) WDI 98% LEB Life expectancy at birth, total (years) WDI 98% Income/c Gdp/c-PPP GDP per capita, PPP (constant 2000 international $) WDI 93% (9%r) Volatility of GDP Maximum value of mobile standard deviation Computed 93% of GDP/c over 5 years period Gini Gini of distribution of h.income per/c WIID2 70% (14%r) normalized in terms of gross income/c Δ gini change Change in Gini > 4 points, (0 otherwise) over 5 years periods WIID2 65% (10%r) R Unemployment unemployment rate WDI 33% (39%R) Prices Inflation consumer prices (annual % change) WDI 73% CPI Consumer price index (2000 = 100) WDI 83% (9%r) Deflator of GDP GDP price deflator index (base year varies by country) WDI 93% (9%r) Cvm_dgdp Price instability (5 years mobile coefficient of variation of gdp deflator) Computed 93% (9%r) R Food_price Food price index (2000 = 100) WDI 48% R Food price/CPI Relative food price (ratio of food price index Derived 45% to overall consumer price index) Education Female illiteracy female illiteracy rate (% of illiterate female aged 25 and above) Barro-Lee 87% (8%r) Female 1ary educ female primary complete (% of adult female ages 15 and above) Barro-Lee 62% Female 2ary educ female secondary complete (% of adult female ages 25 and above) Barro-Lee 62% Fem post2ary ed. female post secondary school complete (% of adult female ages 25+) Barro-Lee 63% Health Res. PHE/ GDP Health expenditure/GDP (normalized into WDI,OECD,GFS 75% general public health expenditure) Physicians/1000 Physicians (per 1,000 people) WDI 85% (8%) DPT immunisation % of children 12-23 months immunized UNICEF/WHO 95% (44%r) against diphtheria, pertussis, tetanus R Antenatal_care % of pregnant women 15-49 checked at least once by a skilled medic UNICEF/WHO 19% (25%R) R Safe births % of births attended by skilled health staff UNICEF/WHO 27% (27%R) R Sanitation Percentage of population with access to improved sanitation facilities UNICEF/WHO 32% (37%R) R Clean Water % of population with access to clean water source UNICEF/WHO 35% (38%R) Out_of_pocket Out-of-pocket health expenditure (% of private expenditure on health) WDI 20% Environment CO2 Carbon dioxide emissions (tons per capita) WDI 97% (10%r) R malaria incidence Average annual number. of reported malaria WHO 28% (32%R) cases divided by the total population Demography Fertility Total Fertility rate (average births per woman during their fertile age). WDI-UN-DHS 99% Migration stock Percentage of foreign born residents on total present population UN Pop Div 90% Behaviour Alcohol Alcohol consumption per capita, litres per person aged 15+ WHO 88% (8%r) R Smoking Prevalence of smoking (all tobacco products) IMASS 20% (91%R) among males 20-55 years old. Shocks HIV-AIDS HIV Adult (15-49 %) prevalence rate USCB, UNAIDS 100% Disasters dummy disasters (=1 if 5 or more % population CRED 100% affected over 5 yrs considered) War weighted number of people killed during episodes of war CRED 100%

Source: authors’ compilation. Notes: r = percentage of data interpolated or included on the basis of our best judgement; R = regional, i.e. a variable that has relevance only at the regional level.

56 Globalization and Health Knowledge Network Table 19: Overlap between geographical and cluster classifications of country groupings

Regions identified on the basis of cluster analysis Region2 HIGH LOW MIDDLE TRANS Unassigned Total

Regions HIGH 20 0 0 3 0 23 aggregated LOW 0 35 6 1 1 43 on the basis MIDDLE 1 2 32 8 2 45 of TRANS 0 0 9 15 1 25 geographical criteria Total 21 37 47 27 4 136

Source: authors’ calculations on GHND. basis of other information at our disposal. The number To check the robustness of this region-based clas- of data points estimated in this way represents no more sification we carried out a k-mean cluster analysis, than 6 percent of the total number of observations in- where the clusters are identified on the basis of four cluded in the database, the exception being DPT vac- high-frequency variables which characterise the cination, for which expert information on the stabil- economic-health nexus in many countries, i.e.: the ity of vaccination at low coverage during the 1960-80 U5MR, the log of GDP/c, the total fertility rate and period allowed us to add missing data, which accounts the normalized Gini coefficient of income distribu- for 18 percent of the data on vaccination12 used in the tion. The results of the analysis (Table 19) show con- regression analysis. The GHND database, and the sup- siderable overlap between the clusters and the four porting documentation, is available at http://www.dse. regional aggregations identified above. 34 out of unifi.it/sviluppo/database_eng.html. the 136 countries (3 from the high income group, 8 from low income, 13 middle income and 10 transi- Most of these variables are relevant to all countries, tion economies – or 13, 16, 24 and 36 percent of while others (denoted by an R in Table 18) are rel- their respective country groups) appear however to evant only to some regions. The 136 countries includ- be ‘misclassified’ in the original regional groups, and ed in GHND are organized into eight geographical could be reassigned to the new groups identified on regions (Sub-Saharan Africa, South Asia, South East the basis of the cluster analysis. This was not done, Asia, MENA, Latin America, OECD, Eastern Eu- however, as it was thought that the parameters of the rope, and the former USSR). These have then been regional regressions (see section 6.5) would not be grouped into four regions which, as suggested earlier substantially affected. in the paper, are characterized by different mortality patterns and are likely to be affected by globalisation 6.2 Descriptive statistics and bivariate correlation through different pathways. The first group, the ‘low between main variables income developing countries’, includes Sub-Saharan Africa and South Asia. The second (‘middle income Here we present the descriptive statistics for the vari- developing countries’) comprises all South East Asian, ables retained in the regression analysis. Some relevant Latin American and MENA countries, while the third variables (water and sanitation coverage, out-of pock- includes the ‘transition economies’ of Central Europe et costs as a share of total health expenditure, and so and the former USSR. The fourth group of ‘devel- on) were dropped because of their low data coverage oped nations’ includes all OECD countries. (Table 18). As they refer to very different countries

12 The vaccination data imputed in the database represent 44 percent of all DPT data, but only 18 percent was used in regression analysis. Not using this 18 percent of data would have prevented the use of all information for the years 1960-75. In any case, the descriptive statistics before and after imputation hardly change.

Pathways of transmission and evidence of impact 57 Table 20: Descriptive statistics for the variables used in the regression analysis

Variables Min Mean Max LEB 27.43 61.82 81.74 IMR 2.60 65.68 285.00 U5MR 3.00 97.33 500.00 GDP/c_current US $ 41.38 3867.68 54741.99 Gdp/c_constant ppp $ 348.99 6683.60 36429.91 Log of Gdp/c constant ppp $ 5.86 8.27 10.50 Time Volatility of Gdp/c 2.52 282.95 4210.29 Gini coeff of income inequality 19.17 42.67 78.15 Inflation (GDP deflator) -3.55 36.44 5085.90 Consumer Price Index 0.00 48.92 1464.30 Food price index 0.00 73.45 2238.43 Fertility rate 0.91 4.09 8.50 Migration stock 0.00 90.33 3145.08 Unemployment 0.55 9.34 43.50 PHE/ GDP 0.01 2.92 9.52 Physicians per 1000 people 0.01 1.06 9.67 Log Physicians/ Log GDP 0.29 0.73 1.22 Immunisation against DPT 1.00 61.48 99.00 Illiteracy among females over 25 0.00 38.22 99.90 Females > 25 with completed 1ary 0.00 15.27 64.80 Females > 25 with completed 2ary 0.00 8.23 49.10 Females > 25 with comp post 2ary 0.00 3.03 27.70 Illiteracy among females over 15 0.00 37.11 99.70 Females > 15 with completed 1ary 0.00 14.85 64.10 Females > 15 with completed 2ary 0.00 8.24 57.00 Alcohol consumption (liters/c) 0.00 5.59 25.31 Smoking incidence 19.00 51.90 90.34 Obesity 1.90 11.32 30.60 HIV-AIDS incidence 0.00 0.91 38.04 Disasters 0.00 0.15 1.00 War mortality 0.00 0.00 0.22

Source: authors’ compilation on the basis of GHND

and a long time span, the variables exhibit consider- 6.3 Model specification and regression plan able variance. However this variance decreases when the data for each of the four main regions (high, As noted above, the GHND is organized as a tri-di- middle, low income and transitional countries) are mensional matrix, with 136 countries on one axis, ten analysed separately, since they are composed of much quinquennia on the second and the relevant variables more similar countries. In this case, the inter-country on the third. The dataset is particularly suitable for variance is lower, though the time variance remains estimating relations using panel data models which considerable. take into account the fact that the sample comprises

58 Globalization and Health Knowledge Network a certain number of countries observed in several pe- to capture simultaneously the impact of the variables riods over time. The linear regression model takes the included in the five mortality models illustrated above following form: (material deprivation, technical progress in health, acute stress, lifestyles and social fragmentation pathways) as

yit = a + xit • b + ui + eit well as the effects of the random shocks discussed in section 5. This multi-pathway approach is considered where y is the dependent variable, x is a vector of explan- appropriate, given the large number of very different atory variables, the subscripts i and t represent respec- countries and long time span on which the estimation tively the countries and the quinquennia of the panel, is carried out. The same relation is then estimated on ui is the error term for each country, eit is a joint error four regional aggregates, i.e. for the high, medium and term for countries and time periods, and a and b are low income countries, and for the economies in transi- the model’s parameters to be estimated. Given the nation. In the estimation of the regional models, the vari- ture of our dataset, the standard OLS procedure yields ables which do not influence mortality in the specific inefficient estimates and causes distortions in the values regions considered (e.g. the relative price of food in of a and b, as information on the countries’ fixed effect the advanced countries, or HIV-AIDS in the econo- would be neglected (Baltagi 2006, Wooldridge 2001). mies in transition) are omitted ex-ante, while in some The estimation procedure best suited to our data, or to cases (such as the migrant stock/resident population) it all situations in which ui is a deterministic parameter is possible that the expected sign of the variable varies that varies from country to country or region to region from region to region. In addition, when there were too included in the panel, is the ‘fixed effects model’.13 We few observations, some variables were omitted from the have not opted for a ‘random effects model’ as this is regional models. best suited in those cases in which ui is a random variable with zero mean and constant variance, i.e. when dealing The estimation procedure adopted follows the logic with a sample that, while representative, does not repro- of a stepwise regression. First we include among the duce all aspects of the population, as is normally the case regressors those explanatory variables which are most when dealing with a sample of households or firms. The correlated and which have been shown in the litera- Hausman test confirmed that the fixed effect model is ture to affect health status. The next most correlated the most appropriate under these circumstances, and is and important variables are then added and so on. As preferable to a random effect model. shown in Table 21 below, examination of the matrix of bilateral correlation coefficients indicated a strong In addition, the estimation procedure adopted tries correlation between explanatory variables which are to take into account the fact that the panel is unbal- all strongly correlated with the mortality indicators, anced, due to a non-negligible number of missing val- as in the case of the total fertility rate. These were ues, (especially for 1960, 1965 and 2005), so as to therefore dropped to avoid serious multi-collinearity reduce the problems of inefficiency and inconsistency problems. of the estimates. This procedure, however, cannot prevent the elimination of the years for which not all data A few explanatory variables were either dropped or are available, a fact that reduces the number of obser- normalized in order to improve the fit, improve the ro- vations used in the estimation of the parameters, par- bustness of the estimates and avoid multi-collinearity ticularly for regions such as Sub-Saharan Africa and problems. For instance, the Gini coefficient14 was in the transition economies before 1990. some cases standardized for the time trend (proxied by ‘year- 1959’) to remove its inverse correlation with The results presented below refer to the estimation of a GDP/C. The variable ‘log doctors per 1000 people’ was global relation between health status and all its determi- divided by the Gini coefficient or by the log of GDP/c, nants, and the parameters estimated in this way ought obtaining an index of availability of health personnel

13 The estimates of the fixed effects model include an intercept for each of the 136 countries in the GHND panel. Such intercepts capture specific country effects due to geography, institutions and unobservables. For this reason, adding a dummy for tropical countries would cause problems of multi-collinearity. 14 Gini coefficients are taken from WIDER’s World Income Inequality Database, the most complete and updated dataset of high-quality and well documented data on income inequality (http://www.wider.unu.edu/wiid/wiid.htm).

Pathways of transmission and evidence of impact 59 Table 21: Correlation matrix of selected variables included in estimation of the global model

LEB IMR U5MR Gini Gini* lnphy ΔGini Migrat Smoking DPT Fem Alcohol HIV- war GDP/c Dis lngdp > 4 stock imm illiter. Cons./c AIDS mort. volat LEB 1.00 IMR -0.92 1.00 U5MR -0.93 0.99 1.00 Gini -0.52 0.49 0.48 1.00 income Gini*log -0.07 0.05 0.03 0.84 1.00 Gdp/c Lnphys/ 0.62 -0.53 -0.56 -0.42 -0.18 1.00 1000 lnGDP/c Δ Gini >4 -0.14 0.15 0.15 0.10 0.00 -0.13 1.00 Migration 0.17 -0.16 -0.16 -0.10 0.00 0.14 0.01 1.00 stock Smoking -0.43 0.35 0.36 0.11 -0.03 0.00 0.01 -0.12 1.00 DPT 0.68 -0.76 -0.76 -0.26 0.07 0.40 -0.01 0.11 -0.29 1.00 Immunisat. Fem. -0.78 0.86 0.85 0.40 -0.01 -0.48 0.14 -0.07 -0.03 -0.68 1.00 illiteracy Alcohol 0.44 -0.49 -0.47 -0.37 -0.08 0.27 -0.06 -0.04 -0.08 0.37 -0.60 1.00 cons. HIV-AIDS -0.43 0.17 0.17 0.31 0.22 -0.30 0.01 -0.06 …. 0.10 0.04 -0.04 1.00 War 0.07 -0.07 -0.06 0.01 0.06 0.03 -0.02 0.24 -0.08 0.07 -0.03 -0.05 -0.01 1.00 mortality GDP/C 0.54 -0.56 -0.53 -0.35 -0.03 0.26 -0.03 0.14 -0.53 0.50 -0.48 0.37 -0.07 0.07 1.00 volatility Disasters -0.20 0.16 0.15 0.15 0.04 -0.13 0.02 -0.10 -0.10 -0.07 0.07 -0.17 0.20 -0.02 -0.13 1.00

Source: authors’ calculations on GHND.

relative to the GDP/c norm. Output volatility was twenty-five years) but not in the other regions or in the proxied by the maximum value in any five year period global model, as it is very heavily trended and therefore of the five year rolling standard deviation of GDP per correlates closely with GDP/c. capita. The health impact of technical progress was proxied in two ways. First, the coverage of immuniza- 6.4 Global results tion against diphtheria, pertussis and tetanus (DPT) was used as a proxy for overall immunization coverage, Table 22 presents the results of the estimation of the oral rehydration therapy and delivery care. In addition, global relation between health status (measured alter- regional time dummies were introduced for the period natively by LEB, IMR, U5MR) and its determinants. 1985-2005 to capture the effect of health progress dur- The parameters in Table 22 capture the average global ing the recent Globalisation Era compared to the pe- impact of the explanatory variables included in the riod 1960-80. All other variables were included with- five mortality models analyzed, and of the random out changes in their usual metric. The age dependency shocks discussed in section 4. Given the large num- ratio was introduced only in the transition economies ber of very different countries and long time span on (where this variable has changed markedly over the last which the estimation is based, the estimates of this

60 Globalization and Health Knowledge Network Table 22: Results of worldwide regression analysis for 1960-2005 on LEBa, IMR, U5MR

LEB IMR U5MR 1960-2005 1960-1980 1980-2005 1960-2005 1960-2005 (1) (2) (3) (4) (5) Constant term 38.966*** 52.707*** 39.673*** 165.088*** 254.926*** Dummy tech progr 0.792* …. …. 3.818 8.297** WS 1980-05 OECD Dummy tech progr 1.362** …. …. 1.192 1.051 1980-05 E.Asia Dummy tech progr -2.461*** …. …. -3.064 1.2714 1980-05 Trans Dummy tech progr 3.311*** …. …. -21.126*** -31.583*** 1980-05 LAC Dummy tech progr 3.432*** …. …. -44.478*** -56.951*** 1980-05 MENA Dummy tech progr 3.397*** …. …. -8.411** -25.279*** 1980-05 S.Asia Dummy tech progr 1980-05 SSA 2.438** …. …. -12.971*** -18.3312***

Log GDP/c 3.203*** 2.307*** 3.148*** -14.19*** -21.75*** GDP/c volatility -.0009** -.0007 -.0008* .0042* .0093*** Gini income distribution -0.057** -.1058** -.0498*** .3215*** .4424** Δ Gini coeff > 4 points -.0423*- -.0645 -.0398 .0861 .2821 Female illiteracy (%) -.098*** -.2763*** -.0427* .5779*** .9464*** Log physicians per 36.89*** 7.305 55.392*- -90.98*- -158.69* 1000 people /Gini DPT Immunisation rate (%) .0861*** .1425*** .0828*** -.3631*** -.6155*** Immigrants stock/ .0026*** .0040*** .0042*- -.0007 -.0042 Total population Alcohol consumption/c -.2536*** -.4074*** -.2702*** .4841* .3820

War and humanitarian 14.95**WS -24.420 13.56*-WS -15.63 -4.332 emergencies Disasters .2864 .4415 .2106 2.6132* 3.2774*- HIV/AIDS -.8495*** -2.099*** -.7737*** 1.1505*** 1.8334***

F statistic 126.89*** 56.45*** 77.02*** 113.39*** 113.64*** R square .897 .847 .890 .820 .845 Number of observations 556 234 385 556 553 Number of countries 97 65 97 97 97

Source: authors’ calculations based on GHND. Notes: *** significant at the 1% level; ** between 1 and 5% level; * between 5 and 10 % level; *- significant at 10-15%. ws = wrong sign. a/ The use of LEB (instead of 100-LEB, as in Table 1) does not change the results of the model, as every linear transformation of the dependent variable in a regression model does not change the value of the parameters (which however take the opposite sign) obtained when using LEB as dependent variable.

Pathways of transmission and evidence of impact 61 Figure 13: Elasticities of the explanatory variables’ coefficients on global estimates for LEB

Elas tic itie s fo r g lo bal m o de l

log GDP/c 0,421 gini income distr. -0,041 ln(physicians)/gini 0,006 Delta gini>4 0,000 Immigrant/pop 0,004 immunization 0,088 female illiteracy -0,046 alcohol -0,023 hiv/aids -0,018 w ar and 0,000 volatility of gdp -0,006 disasters 0,001

-0,15 -0,05 0,05 0,15 0,25 0,35 0,45

Source: authors’ calculations based on GHND. Note: the elasticities are calculated at the mean value of each variable.

‘multi-pathway model’ are quite stable. The regression nical progress in health’. The negative value recorded results on LEB are presented separately for the entire by this variable in the transition economies is highly 1960-2005 period, and the sub periods 1960-80 and plausible as it reflects the dismantling of the socialist 1985-2005. Models (1), (4) and (5) present the re- health care model in the 1990s and the difficulties met sults of the estimation of a multi-pathway mortality in replacing it with a new system. It should be noted, model over the 1960-2005 period. Though based on however, that ‘technical progress in health’ is proxied by 556 observations (553 for U5MR) out of a theoreti- a time dummy that might capture other unexplained cal maximum of 1360 for 97 countries (out of 136), effects such as – in the case of the transition econo- the estimates in Table 22 are very satisfactory. Practi- mies - the shock experienced after 1990. The impact cally all variables have the expected sign and plausible of technical progress in health is captured also by the and statistically significant coefficients.15 To render the significant values of the coefficient of DPT vaccination. impact of the explanatory variables on LEB (model 1 For instance, model (5) suggests that raising the DPT in table 22) easily comparable, Figure 13 reports their immunization by 30 points would reduce U5MR by elasticities except for the dummy technical progress. 18.4 points. Log GDP/c is highly significant for the 1960-2005 period and for the two sub-periods consid- Models (1), (4) and (5) in Table 22 suggest that an im- ered. The same is true for the volatility of GDP/c that portant part of the gains in LEB, IMR and U5MR (rang- affects negatively, if modestly, LEB, IMR and U5MR. ing between 0.74 years in the OECD to 3.4 LEB years In turn, income inequality affects strongly and signifi- in South Asia and MENA) realized in the 1985-2005 cantly all three health indicators (see Figure 14 for its period compared to 1960-1980 was due to the ‘tech- relationship with LEB).

15 This does not exclude however the possibility of reverse causation which is usually tested by means of the Granger test. However, this test is not suitable for the GHND quinquennial dataset in which each variable has at best ten observations (Hurlin and Venet 2001). It is therefore more appropriate to deal with this problem in theoretical terms. In this regard, it should be noted that reverse causality makes no sense in the majority of the relations in Table 22 (for instance it is not plausible that an increase in IMR can raise economic volatility, or that a fall in LEB will rise the interperiodal Gini variations larger than four points). The only relation in which reverse causation may be plausible is that between LEB and GDP/c. In this case, however, the relation between rising LEB (due for instance to a fall in U5MR) and higher GDP/c would be characterized by time lags, thus excluding the possibility of reverse causation on sychronous data. The parameters are also affected by estimation bias caused by the omission in the regression analyses of a few variables discussed in section 4 - such as out-of pocket health costs and coverage of health insurance which were dropped because of insufficient data. In addition, the parameters may be further distorted by the possible endogeneity of some explanatory variables, which are simultaneously determined by the dependent and policy variables. Solving formally this endogeneity problem by means of a simultaneous equations system is however a difficult task in a panel with 136 countries.

62 Globalization and Health Knowledge Network Figure 14: Relation between inequality and LEB, 1960-80 and 1980-2005

0 1

8 0

6 0

4 0

2 0 2 4 6 8 2 4 6 8 0 0 0 Gini0 nor0m 0 0 0

leb Fitted values

Graphs b y time 2

Source: authors’ elaboration based on GHND – On the vertical axis, the dots are the observed and the line is the fitted LEB. Gininorm refers to the normalized Gini.

Increases in income inequality greater than 4 Gini points health behaviors – such as excessive alcohol consump- from one quinquennium to the next were also found to tion – affects LEB and, surprisingly, IMR. Finally, the be significant in all three LEB models, although their model confirms a positive, if small, impact of migrant elasticity is very small and they are – as expected – stocks greater than 3.5 percent of the resident popula- weakly significant in the case of IMR and U5MR. For tion on the LEB of the countries of destination. This instance, a 10 points rise in the Gini coefficient from result is interesting in the light of the current debate on one quinquennium to the next reduces at the margin the impact of migration. LEB by one year, strengthening the conclusions of prior studies about the effect of sudden shifts in inequality on As for the ‘random shocks’, HIV-AIDS appears to have cardiovascular and violent mortality. Female illiteracy is a large and significant effect on LEB, IMR and U5MR. strongly significant in all five models, confirming the Raising its prevalence rate by 30 points reduces LEB by findings of the micro-studies cited in section 4. For in- a staggering 25 years (as observed in Botswana), and stance, a reduction by ten percentage points in female raises child mortality by 52 points, i.e. values similar to illiteracy raises average LEB by 0.76 years, and reduces those estimated in Cornia and Zagonari (2002). ‘Disas- IMR and U5MR by 5.1 and 8.2 points respectively. ters’ and ‘war and humanitarian emergencies’ are non significant and often have the wrong sign. This may be The availability of health services (measured by log due to their low frequency (only 87 and 63 have non- physicians per 1000/ Gini) is significant in all five zero values out of 556 observations) or to low cover- models in Table 22, but is weakly significant for the age of the CRED database, or because the variable was 1960-80 period. In turn, adopting health-damaging poorly proxied in the GHND.16

16 A closer look at the variable ‘wars and humanitarian emergencies’ suggests that its non significance may be due to low number of non-zero data points in GHND, i.e. 64 out of 583. As ‘wars’ have been more common among low and middle income countries, we tested the significance of this variable for these regions separately but, also in this case, the results were with the wrong sign and non significant. And also in this case, ‘wars’ seemed to be an uncommon phenomenon (only 20 percent of the cells were non-zero). Only in the middle income countries affected by conflicts (Lebanon, Guatemala, Iran, El Salvador, Philippines, Algeria, Colombia, Viet Nam, and 15 others) did the variable take the right sign but is non significant. It would thus appear that – while wars do increase sharply mortality in places of large conflicts such as Viet Nam and Lebanon – their impact is comparatively small compared to the impact of mortality due to traditional structural factors.

Pathways of transmission and evidence of impact 63 Figure 15: Observed and fitted median spline of world LEB

Source: authors’ calculations based on GHND.

Figure 15 shows (solid line) the observed trend in the While log GDP/c captures unambiguously the impor- LEB spline (the unweighted median of LEB for all 136 tance of resources for health in the ‘material depriva- countries included in the panel) which, as discussed tion pathway’, income inequality captures simultane- in section 2, grew more slowly from the early 1980s ously the effects described in the ‘psychosocial stress’, onwards than during the previous twenty years. It also ‘material deprivation’ and ‘social cohesion’ models. shows that the LEB predicted by model (1) (dotted line) fits very well the actual trend except, perhaps, for How stable over time is the relation estimated in col- the first 5 years. A comparison of the stability of the umn (1) of Table 22? Columns (2) and (3) of the same estimates in models (1), (2) and (3) shows that while table provide estimates of the parameters of the socio- the sign and size of the parameters remained broadly economic determinants of LEB over 1960-1980 and unchanged, the significance of a few parameters was 1985-2005. These results show that while the over- lower over 1960-80, for example in the case of the vol- all significance of the estimates (F test) remains very atility of GDP/c and alcohol consumption, possibly high and the sign and size of the parameters remained because these two phenomena were less pronounced unchanged, their statistical significance was reduced over the years 1960-80. It also shows that female il- in some cases. This is normal as most variables which literacy became non significant over 1985-2005 as it show lower significance (column 2) concern phenom- became a less common phenomenon. The loss of sig- ena such as the volatility of GDP/c and were less pro- nificance is also due to the smaller number of observa- nounced during the 1960-80 period. tions and reduced variance of the variables in each of the two sub-periods. From a theoretical perspective, it should be noted that the above regression model includes variables which To conclude, it should be noted that the reduced form are part of all five pathways discussed in section 4, as models in Table 22 contain variables included in more well as a few ‘random shocks’ (AIDS, wars and disas- than one of the five models discussed in section 4, and ters). Variables such as log GDP, log doctor per 1000/ that in these cases the impact estimated by model (1) Gini, female illiteracy and, possibly, migrant stock is the sum of the effects of different mortality models. emphasize the importance of ‘resources for health’ as

64 Globalization and Health Knowledge Network Table 23: Results of the regression analysis on LEB, IMR, U5MR for high and middle income countries, 1960-2005

High income countries (5 Middle inc. countries (5 quinquennial data) quinquennial data) LEB IMR U5MR LEB IMR U5MR 1960-05 1960-05 1960-05 1960-05 1960-05 1960-05 Constant term 94.43*** 2.5942*- -5.4486 38.56*** 158.22*** 208.53*** Dummy med progr 1.06*** -3.9686*** -5.1890*** 80-05OECD Dummy med progr 0.6603 9.19**ws 12.49*-ws 80-05 E.Asia Dummy med progr 1.8315*** -13.40*** -19.07*** 80-05 LAC Dummy med progr 1.2796* -22.96*** -34.55*** 80-05MENA

Log GDP/c ………… ………… 2.5239*** -13.31*** -18.37*** Log GDPc / volatility 74.78*** -30.22 -102.16 GDP/c volatility …………… … ……… -.00003 .0071 .0169* Gini income distribution -.0482* .3415*** .5013*** Gini income distribution / t …………… …………… -.0679*** .2534*** .4695*** Δ Gini coeff > 4 points .1002 -.5582 -.7957 -.0456 .1780 .6743*- % women>25 with .2888*** -.7689*** -.9792*** …………… …………. post 2ary educ …………… Female illiteracy …………… …………… -.1918*** 1.1433*** 1.7979*** Log phys. per 1000/ -28.9***ws 20.80 36.57 11.2796*** -14.75 -31.33 Log GDP/c DPT Immunisation rate (%) .0775*** -.1969*** -.2514*** .0701*** -.3296*** -.5999*** Immigrants stock/ .0017*** .0029*-ws .0027 .0035*- .0099 .0128 Population Alcohol consumption/c -.2582*** .4301** .5003*** -.4350*** 1.4932*** 1.7121** Smoking -.0544*** -0.0210 -.0306 No data No data No data

War and human emergencies Unobserved Unobserved Unobserved -10.3808 94.22 136.06 Disasters Unobserved Unobserved Unobserved -.2075 1.6205 1.3079 HIV/AIDS Unobserved Unobserved Unobserved Unobserved Unobserved Unobserved

F statistic 69.81*** 40.97*** 40.85*** 185.63*** 115.03*** 127.17*** R square .287 .509 .513 .837 .684 .721 Number of observations 130 130 130 212 212 212 Number of countries 22 22 22 34 34 34

Source: authors’ calculations Notes: *** significant at the 1% level; ** between 1 and 5% level; * between 5 and 10 % level; *- significant at 10-15%. ws = wrong sign (i.e. a sign different from that expected on the basis of the theories discussed earlier)

Pathways of transmission and evidence of impact 65 posited by the ‘material deprivation pathway’. In turn, Smoking and drinking depress LEB in a significant way, the time-dummies and DPT coverage emphasize the while immigration has a positive if modest effect on contribution of technical progress in the medical field. LEB. As predicted by the Preston curve (1976), GDP/c Large increases in Gini as well as high output volatility is not significant if considered alone, but turned out to reflect the impact of psychosocial stress, while alco- be significant when interacted with its volatility. Like- hol consumption per capita and (in regional models) wise, increases in Gini coefficients > 4 are not signifi- smoking incidence typically characterize the ‘health cant as this phenomenon was seldom observed in the behaviour pathway’. Finally, income inequality may region. All other variables have correct signs and plau- capture the effects posited by the ‘social cohesion and sible parameters. In the case of IMR-U5MR, half of the hierarchy models’, but it also relevant for the first two variables have the right sign and are statistically signifi- pathways (resources and stress). cant, although the economic variables measuring stress and resources (log GDP/volatility, income inequality 6.5 Regional results and sudden changes in inequality) have non significant signs. As expected, smoking and migrant stock are not The mortality models described in Table 22 were fur- significant suggesting they do not significantly affect ther tested on the high-income, medium-income, the survival of children but do affect that of adults. low-income, and transition-economies to see if the intensity of the relations described above holds also The mortality model was also validated in a satisfac- in the separate country groups. The variables that do tory way for the middle income countries (218 ob- not influence mortality in each specific region were servations), as all twelve variables (including war and omitted (e.g. the relative price of food, wars and di- disasters) have the expected sign and are highly signifi- sasters in the advanced countries, or HIV-AIDS in the cant with the exception of Δ Gini > 4, GDP volatility, economies in transition), while the expected sign of disasters, wars and medical progress in East Asia. Sim- some variables (e.g. the migrant stock/resident popu- ilar results are obtained for IMR and U5MR. Interest- lation) may change from region to region, and that of ingly, the latter rise when the ‘migrant stock’ increases others (e.g. log GDP/c) may become insignificant in (though the parameters are not significantly different the high income group. Also, the same phenomenon from zero). (e.g. female education) was proxied in ways relevant to local conditions, e.g. using the percentage of women For the low income countries (125 observations) the with post-secondary education in the high income LEB model shows very satisfactory results (Table 24) and transition groups of countries, and female illit- as all variables have the expected sign (except for al- eracy in the medium income group. When available, cohol consumption, disasters and wars which are also regional variables were added, such as smoking in the non significant), and are statistically significant- (ex high income group, while variables with no or only cept for Δ Gini > 4), thus confirming the conclusions few regional observations were omitted. Finally, the arrived at on the basis of the global model. It must estimation of models similar to those in Table 22 at be noted, that the value of the parameters (e.g. of log the regional level requires the division of the sample GDP/c, Gini, DPT and so on) is generally greater of 556 data into four sub samples, and leads by defi- than in the global or middle income model confirm- nition to some loss of significance of the parameters, ing the theoretical expectations that changes in these particularly for regions with few data and large mea- variables have a much greater impact in poor than rich surement errors in the variables. This is the case for countries. Furthermore, in this country group, the the transition countries, where lack of sufficient quin- variable disasters takes the correct sign in the case of quennial data meant that the model had to be tested IMR and U5MR though it is still not significant. on annual data for 1980-2005. Finally, the estimates for the transition countries are The results for the high income countries (130 observa- less complete that those for the other three regions, tions) are satisfactory (Table 23). Of the ten variables as the estimation was carried out on yearly data for explaining LEB all but one (log of physicians per 1000/ 1980-2005, and as no information was available for log GDP/c) have the right sign and are significant. immunization, migration, alcohol consumption, war

66 Globalization and Health Knowledge Network Table 24: Results of the regression analysis on LEB, IMR, U5MR for low income and transitional economies, 1960-2005

Low income countries Transition countries 1960-2005(quinquennial data) 1980-2005 (yearly data) LEB IMR U5MR LEB IMR U5MR Constant term 22.10** 158.73*** 291.25*** 54.37*** 80.29*** 67.92*** Dummy TechProgr80-05 3.041*** -19.06*** -33.58*** …… …… …… Dummy Trans 90-05 …… …… …… -.15139 -1.128** -.9778*-

Log GDP/c 3.6640*** -7.8231* -18.91** 1.6664** -2.9997** -2.2838 GDP/c volatility -.0064* .0065 .0396*- -.0012*** 0.0007*- .0009*- Gini income distribution -.2032*** .8179*** 1.5343*** -.0062 -.2843*** ws -.3061*** ws Δ Gini income > 4pts .0641 -.2959 -.7060 -.1421** .3840*** .7762*** % enrolment in …… …… …… .0346*** -.0761*** -.0504*** 2ary education %women with .1913** -.6223* -1.0926* …… …… …… completed 1ary Log physicians x1000 14.2350** -14.48 -24.7722 8.6528** -36.66*** -32.9712*** people / Log GDP/c DPT Immunis rate (%) .1147*** -.5339*** -.8950*** …… …… …… Immigrants stock/ Population -.0276** .0030 .0818 …… …… …… Age dependency ratio …… …… …… -15.2211*** 45.395*** 59.2501*** Alcohol consumption/c -.0240 -.2135 -1.4582 No data No data No data

War 26.66 - 198.94 -291.25 No data No data No data Disasters .0246 4.7663** 7.353* No data No data No data HIV/AIDS -.7208*** .6006*** .9750*** Unobserved Unobserved Unobserved

F statistic 32.21*** 32.60*** 27.49*** 28.95*** 43.66*** 45.36*** R square .743 .661 .613 .170 .586 .605 Number of observations 123 123 114 325 316 316 Number of countries 23 23 22 24 24 24

Notes: the years 1960 and 2005 have a low coverage Notes: *** significant at the 1% level; ** between 1 and 5% level; * between 5 and 10 % level; *- significant at 10-15%. ws = wrong sign (i.e. a sign different from that expected on the basis of the theories discussed earlier)

and disasters. Several of the key effects are, however, cant, though its large increments are. Finally, similar correctly estimated also in this case: model 1 (Table 24) results were obtained for the IMR and U5MR, though captures most effects discussed so far, i.e. those related also in these two models the Gini of income distribu- to the negative progress in medical research, GDP vola- tion had the wrong sign, and log GDP was non signifi- tility, GDP per capita, large Gini increments, female cant. A more complete dataset is needed to test fully the enrolment rate in secondary education, the availability above mortality models for this region. of doctors, and dependency ratio. The latter is a variable which is particularly relevant in this region, since it A confirmation of the good results of the estimations suffered a real demographic collapse in the 1990s. The presented in Tables 23 and 24 is given by a compari- Gini of income distribution is statistically non signifi- son of the LEB elasticity of the same variable in the

Pathways of transmission and evidence of impact 67 Figure 16: LEB elasticities of explanatory variables for four country groups

Source: authors’ calculations based on GHND. Note: the elasticities are calculated at the mean value of the variables.

68 Globalization and Health Knowledge Network Figure 17: Trend in the observed (solid line) and estimated (dotted line) of the median spline of the life expectancy at birth in the four sub regions

four different regions (Figure 16). In fact, most of the low income countries, while large changes in Gini such elasticities confirm the predictions of the theo- between periods are relevant only in the economies retical models discussed in section 4. For instance, the in transition. Immigration has a negative LEB elas- LEB elasticity of GDP/c is highest in the low income ticity in low income countries, but a positive one in countries (where the material deprivation model is the middle and high income ones. The availability of dominant) and lowest in the high income ones (where physicians has a fairly high LEB elasticity in all regions instead the highest LEB elasticities are found for in- but in the advanced countries. Finally, the good fit of come inequality, smoking and alcohol consumption, the regional estimates (dotted lines) is confirmed also as suggested by the ‘lifestyles’ and ‘inequality and social in relation to the observed values (solid line) for the cohesions paradigms). Likewise, as expected by theory, respective reference periods and four regional groups volatility and immunisation have a greater impact in (Figure 17).

Pathways of transmission and evidence of impact 69 Globalisation and Health: Pathways of Transmission and Evidence of Impact

7. Simulation of LEB changes due to globalisation & shocks

he above global and regional models can be used • log of GDP/c*Gini, log of physicians per to assess whether the changes in the policy-driv- 1000/log GDP/c, and migration stock, rose Ten health determinants intervened in the recent Glo- over 1985-2000 in line with their 1960-80 balisation Era in the policy-driven health determinants trend; (growth of GDP/c, inequality, instability, health provi- • DPT immunization rates, female illiteracy sion, and so on), random shocks (HIV-AIDS, wars and (or primary or secondary education) and disasters) and endogenous medical progress have led by alcohol consumption rose in line with their 2000 to LEB values which were higher, equal or lower 1960-90 trend; than those which would have been achieved under a • GDP volatility, the Gini index of the distri- ‘counterfactual scenario’ in which the determinants of bution of income, age dependency ratio, and LEB did not change their 1980 values throughout the smoking incidence, remained at their 1980 1980s and 1990s, or evolved over these two decades ac- or 1985 level. This means that – as a conse- cording to the time trend of 1960-80 or 1960-90. quence – the Δ Gini variables take the value of zero; In practice, this meant fixing for each independent • there was no progress in health technologies variable a 2000 value obtained by prolonging over over 1985-2000 time the 1960-80 (or 1960-90) trend or keeping con- • HIV-AIDS incidence remained at its 1980 stant its 1980 value and, second, simulating, on the level and there were no disasters and wars. basis of these counterfactual values and the parameters of the regional models in Tables 23 and 24, the values Thus, for each region, the simulated values LEB in s LEB would have taken by 2000. The specific assump- 2000, indicated as LEB i are equal to the sum of tions made in the counterfactual scenario are that: the products of the simulated 2000 values of the

70 Globalization and Health Knowledge Network explanatory variables by the parameters a1, a2, etc of At the global level, the policy-driven changes appear to equations (1) in Tables 23 and 24. In symbols: have reduced LEB by 1.52 years as a result of several offsetting effects. Higher income inequality than in the s s LEB i = a0 + a1 Dummy Health Progr (=0) + a2 Log counterfactual scenario depressed LEB by 0.77 years. s s s GDP/c * Gini - a3 GDPvolatility - a4 Gini - a5 This loss was counterbalanced by LEB gains (0.73 s s ΔGini>4 - a6 Female Illiteracy + a7 Log Phys per years) due to a growth in GDP/c faster than in the s s s 1000/Log GDP/c + a8 DPT + a9 Migrant stock - counterfactual scenario in China, India and the rest of s s a10 Alcohol/c - a11AIDS South Asia (in most regions, however, a GDP growth slower than over 1960-80 contributed to a decline in In turn, the observed values of LEB in 2000, indicated LEB). Smaller LEB losses (0.08 years) were due to o as LEB i, are the sum of the products of the observed large intra-period rise in income inequality, while 2000 values of the explanatory variables multiplied by GDP/c volatility appears to have caused perceptible the parameters a1, a2, etc. of equations (1) in Tables LEB losses in most regions. A rise during the 1980s- 23 and 24. In symbols: 90s in the number of physicians per 1000 normalized by log GDP/c was slower than in the counterfactual o o LEB i = a0 + a1 Dummy Health Progr (=1) + a2 Log scenario in several regions, and led to a global LEB loss o o o GDP/c * Gini - a3 GDPvolatility - a4 Gini - a5 of 0.51 years. Smaller effects were observed for illit- o o ΔGini>5 - a6 Female Illit + a7 Log Phys per 1000/ eracy while improvements in health behavior (alcohol o o o Log GDP/c + a8 DPT + a9 Migrant stock - a10 consumption and cigarette smoking) in the OECD o o Alcohol/c - a11 AIDS and a faster rise in the migrant stock faster generated small but telling improvements in world LEB. To these It is now possible to compute, region by region, the ‘policy-driven effects’ one has to add the LEB changes difference between LEBoi and LEB si as the sum of due to endogenous progress in medical technology and the differences between the observed and simulated (for the transition economies) in age dependency ratios values of each explanatory variable multiplied by the and subtract those due to AIDS, wars and disasters. In- parameters a1, a2, etc of equation (1) in Tables 23 and terestingly, at the global level, the gains due to medical 24, i.e.: progress cancel out almost completely the LEB losses due to policy-driven changes and random shocks. One o s o LEB - LEB = a1 [Dummy Health Progr (=1) - Dum- thus wonders how large these gains would have been s o my Health Progr (=0)] + a2 [Log GDP/c * Gini - Log had the policies towards the transfer of health technol- s o s GDP/c * Gini] - a3 [GDPvolatility - GDPvolatility ] ogy (such as the TRIP agreement and international aid o s o s - a4 [Gini - Gini ] - a5 [ΔGini>5 - ΔGini>5 ] - a6 to high impact health programs) been more favorable. o s [Female Illit – Female Illit ] + a7 [Log Phys per 1000/ A related disturbing message of this simple simulation o s Log GDP/c - Log Phys per 1000/Log GDP/c ] + a8 is that a perceptible LEB loss (0.47 years) was due to o s o s [DPT - DPT ] + a9 [Migrant stock - Migrant stock ] the stagnation or decline in DPT vaccination coverage o s o s - a10[Alcohol/c - Alcohol/c ]– a11 [AIDS - AIDS ] since 1990, a result in line with the findings of Cutler et al. (2006) about the decline in national and inter- In this way, for each of the large countries/regions national support to immunization campaigns during identified below in Table 25, it was possible to derive, the 1990s. Finally, given the low value of their param- variable by variable, the LEB gains (+) and losses (-) eters, the variables wars and disasters do not appear to in the year 2000 due to exogenous globalisation poli- have generated perceptible LEB losses, although this, cies affecting the determinants of health, endogenous as noted earlier, may be due to data limitations and changes and random shocks in relation to a ‘business specification problems. as usual counterfactual’. World changes in LEB and in each of its determinants are obtained by weighing As expected, there were winners and losers from the regional LEB changes by their population size. Table policy changes introduced during the recent Globalisa- 25 summarizes the results of this hypothetical simulation Era. The biggest losers of the policy-driven changes tion that seems sufficiently differentiated, and broadly are sub-Saharan Africa and the two regions in transi- in line with the real life changes observed. tion, but other less obvious findings emerge from

Pathways of transmission and evidence of impact 71 Table 25: Gains (+, green color) and losses (-, red color)of LEB years by 2000 due to policy changes, endogenous changes and random shocks during 1980s-1990s

Region OECD TRANS USSR E.Asia China L. Amer MENA India S.Asia SSA WORLD

Policy driven 2.02 -1.78 -3.92 0.49 -3.61 -1.54 2.19 -1.07 -1.59 -5.63 -1.52 LEB changes Log GDP/c 0.00 -0.43 -1.91 -1.22 3.98 -0.80 -2.07 1.71 0.69 -0.99 0.73 Log GDP/c on -0.46 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -0.07 volatility Gini of income -0.03 -0.07 -0.12 0.00 -2.14 0.00 0.00 -1.15 -0.61 -0.45 -0.77 inequality Gini of income 0.00 0.00 0.00 0.00 0.00 -0.01 -0.01 0.00 0.00 0.00 0.00 inequality / (year-1959) GDP/c Volatility 0.00 -0.72 -0.49 -0.05 -1.26 0.01 0.04 -0.63 -0.32 -0.09 -0.44 Intra-period D 0.02 -0.58 -1.60 -0.08 0.00 -0.03 0.00 0.00 0.00 0.14 -0.08 Gini >4 points Log physicians per -0.44 0.02 0.37 1.10 -1.67 0.25 0.73 -0.97 -0.44 -0.60 -0.51 1000/Log GDP/c Log physicians 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 per 1000/Gini Migrant stock/ 0.07 0.00 0.00 0.41 0.00 0.01 0.39 0.00 -0.12 0.06 0.07 population DPT immunization 0.31 0.00 0.00 0.70 -0.73 -0.05 -0.29 -0.18 -0.58 -3.37 -0.47 coverage Female education 0.52 0.00 -0.16 -0.57 -1.78 -1.14 3.41 0.15 -0.21 -0.32 -0.31 Cigarette smoking/c 0.82 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.12 Alcohol consumption/c 1.21 0.00 0.00 0.20 0.00 0.22 -0.01 0.00 0.00 0.00 0.22

Endogenous driven 1.07 0.36 0.35 0.66 3.04 1.83 1.28 3.04 3.04 3.04 2.15 LEB changesa Age dependency ratio 0.00 0.66 0.66 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.05 technical progress 1.07 -0.31 -0.31 0.66 3.04 1.83 1.28 3.04 3.04 3.04 2.10 in health field

Shocks driven 0.00 0.00 0.00 0.00 -0.02 -0.04 -0.05 -0.57 -0.34 -6.36 -0.76 LEB changesa War and humanitarian 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -0.01 0.18 0.02 conflicts Disasters 0.00 0.00 0.00 0.00 -0.02 -0.04 -0.05 -0.02 -0.02 -0.01 -0.02 HIV-AIDS 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -0.54 -0.31 -6.54 -0.76

Total LEB changes 3.08 -1.42 -3.57 1.15 -0.59 0.25 3.42 1.41 1.11 -8.95 -0.13

Source: authors’ calculations based on the parameters of the regional models presented in Tables 23 and 24. Notes: a/ For the distinction between LEB changes driven by policy, endogenous factors and shocks, see footnote 2. b/The results of the above simulation are based on the regional parameters (tables 23 and 24) The results of the simulation based on the global parameters (Table 22) yields similar results for the world as a whole. As for the ten regions, the signs of the variations are similar but the size of the variations are obviously very different , as the parameters of the explanatory variables vary substantially between the global models (which measure an ‘average’ effect ) and the regional models which reflect different regional circumstances.

72 Globalization and Health Knowledge Network this simulation. Among the winners one can count the OECD, East Asia excluding China, a region where LEB losses due to growing inequality, sudden inequality rises and a growth slowdown were offset by improvements in the field of medical staff, alcohol consumption and eradication of female illiteracy. Surprisingly, MENA experienced large policy-driven LEB gains due to a growth in doctors, migrants and - especially - female education which was faster than in the past. South Asia excluding India exhibits marked LEB losses due to a worse-than- expected performance in all economic and social areas, but large gains due to the transfer of health technologies. Interestingly, the two new ‘growth superstars’, i.e. China and India, appear to have suffered a policy-driven loss of LEB, as the effect of a growth faster than in the counterfactual was offset by losses due to a sharp rise in inequality and volatility, as well as poorer than previously recorded performances in reducing female illiteracy, alcohol consumption, available physicians and DPT vaccination. However, in both China and India, progress in health technologies and limited LEB losses due to AIDS meant that the changes that intervened during the recent Globalisation Era were – on balance – more positive than in the counterfactual scenario.

Though plausible and interesting, the conclusions of the above simulation suffer from a few problems. First of all, except for AIDS, wars, disasters and endogenous changes in health technologies and dependency ratios, the above approach attributes all LEB gains and losses to the impact of globalisation policies on growth, inequality, volatility, vaccination coverage, female illiteracy and so on, while it can be argued that other factors (e.g. institutions, external financing and so on) may have been in part responsible for the changes recorded in these social determinants of health. This problem is addressed in part in the next section. Second, although the multi-pathways model used for the simulation is already fairly complex, lack of data prevented the inclusion in the model of a few determinants of health (e.g. diet and smoking) discussed in sections 4 and 5. Thirdly, technical progress in health care is approximated by DPT immunization coverage and a period dummy. The latter captures the part of LEB unexplained by the variables included in the model and, as such, could reflect the impact of omitted variables. Yet, as the number of variables included in the model is fairly large and its fit rather good, this argument may not carry much weight.

Pathways of transmission and evidence of impact 73 Globalisation and Health: Pathways of Transmission and Evidence of Impact

8. Impact of liberalisation-globalisation on the determinants of health

8.1 Introduction

Section 5 argued that during the last 25 years the pol- ticular the impact of WC-type liberalisation-globalisa- icy-controllable social determinants of LEB (GDP/c, tion policy on three key social determinants of health income inequality, GDP/C volatility, female illiteracy, i.e. GDP/c (expressed for convenience as a GDP in- DPT immunisation and so on) improved in most dex, or GDP/c growth rate), the volatility of GDP/c regions less rapidly that during the previous twenty and income inequality. years, while in some cases their absolute level worsened. This unsatisfactory trend was responsible for the 8.2 Methodological problems and data sources ‘policy-driven LEB losses’ recorded between 1980 and 2000 in five of the eight regions analyzed, as well as for An assessment of the effects of policy reform on the world as a whole, in relation to a counterfactual growth, volatility and inequality can be carried out us- scenario that assumed that these social determinants ing four different methodologies: (i) the ‘before-after’ of health improved at the same pace as in 1960-80 approach in which the value of the variable of interest (Table 25). We have argued also on the basis of case is analyzed before and after the introduction of a given studies and theoretical arguments that such changes policy change. This method is normally improved us- were due to a good extent to a widespread, uncritical ing the ‘differences in differences’ approach by which and premature adoption of Washington-Consensus- the before-after changes in the country that introduced type liberalisation policies, which ignored the pecu- the policy change are compared with those observed in liarities of local conditions and institutions. a similar control country in which no policy changes were introduced; (ii) general equilibrium models used In this section, we conduct a formal test to establish to simulate the impact of alternative policy packages; whether the above viewpoint is correct. We test in par- (iii) detailed qualitative ‘structured comparisons’ of

74 Globalization and Health Knowledge Network two (or more) case studies that adopted the reforms financial reforms. Finally, in some cases, e.g. that of to different extents, (iv) cross-country or panel regres- privatisation, the most obvious measures (for instance sions in which the impact variable (growth, inequality, the share of state assets or value added over the respec- volatility, etc) is regressed on some indexes of policy tive national total) conceal more than they reveal, as reform and a number of control variables, as done in privatisation can be carried out in very different ways, this paper. Each of these approaches has some advan- ranging from the well-designed and egalitarian land tages and limitations and, in fact, no ideal tool to as- reform of 1978 in China to the insider privatisation sess the impact of policy reform has been developed of Russia, and to the voucher-privatisation carried out so far. This forces the researcher to use her or his best in the Czech Republic. Although these different ap- judgement on the basis of various forms of informa- proaches have a similar privatisation index, they had tion, including case studies, as done in this paper. different effects on economic performance and equity.

The second methodological issue to be dealt with in this To solve the theoretical and measurement problems type of analysis is that the measurement of ‘liberalisa- presented by the multifaceted nature of many policies, tion’ and ‘globalisation’ policies is, in itself, highly prob- most researchers rely on one synthetic reform index lematic from a theoretical and empirical perspective, summarizing changes carried out in all reform areas. and can give rise to important identification and mea- The computation of such indexes is normally carried surement errors. The textbook model is one in which, by the staff of international organisations, such as re- to enact a policy change, the government has at its gional development banks (for example the IADB, disposal a clearly identifiable policy instrument which ECLAC or EBRD), or global institutions such as the takes a number of finite values. This approach could not IMF and the World Bank. At the IMF for instance, be further away from reality. Even in the comparatively Lane and Milesi-Ferretti (2003) compiled a synthetic straightforward case of ‘trade liberalisation’, policy re- index of international financial integration compris- form can be measured through many different indexes. ing the stock of aggregate foreign assets and liabilities These include policy-based measures such as the aver- as a share of GDP, cross-border equity holdings (de- age tariff rate, tariffs variance and peaks, the value of fined as the sum of the stocks of portfolio equity and tariff revenue on imports, coverage of contingents and FDI assets), and portfolio equity and FDI liabilities. other quantitative restrictions, black market exchange Likewise, Beck et al. (2000) developed a new database rate premium, voluntary export restraint, and social on the size, activity and efficiency of various financial and environmental clauses. In addition, several of these intermediaries. It covers the period 1960-1997 and single indexes can be combined with each other to gen- 175 countries and is based on the IMF’s International erate an overall index, as done for instance in the case of Financial Statistics. The main indexes available from the IMF’s index of trade restrictions, the World Bank’s this database are: central bank assets to total finan- index of trade openness, the Heritage Foundation in- cial assets; bank deposits to total financial assets, bank dex, and the Sachs and Warner index (see later). Finally, deposits versus central bank assets; liquid liabilities some authors use outcome-based measures of trade lib- to GDP; central bank assets to GDP; deposit money eralisation such as the ‘export plus import/GDP ratio’ bank assets to GDP and so on. and the growth rate of exports, however these do not take into account the fact that such outcomes depend Some researchers have produced reform indexes. as much on trade policies as on country characteristics Among them, Sachs and Warner (1995) developed such as size, natural resource endowments, geography, an “openness index”, according to which an economy history, institutions and terms of trade. Among these is closed if at least one of the following conditions measures, the most widely used is ‘exports plus imports is observed: (i) average tariff rates on imports of in- as share of GDP’. termediate and capital goods are 40% or higher; (ii) non-tariff barriers cover 40 % or more of imports of Measurement problems are even more pronounced intermediate and capital goods, (iii) the black market when measuring policy reforms, the implementation exchange rate is depreciated by 20% or more relative of which require changes in institutional and regula- to the official exchange rate, (iv) a socialist economic tory areas, as in the case of the domestic and external system (as defined by Kornai 1992) prevails, and

Pathways of transmission and evidence of impact 75 (v) the state has a monopoly of major exports. This criticized by others (Rodrik, Subramanian and Trebbi, index does not, however, take into account some key 2002), who argue that growth is determined primarily aspects of globalisation such as financial globalisation. by the quality of institutions which, in their words, The index has been computed for 79 countries for the “trumps” everything else, including geography and period 1970-1992, and therefore cannot be readily openness. used for a broader set of countries and years. Some studies have also analyzed the impact of liber- Sectoral and global reform indexes were developed alisation reforms on income distribution. The pro- for the years 1970-1995 by Morley et al. (1999) for ponents of liberalisation and globalisation claim that Latin America and the transition economies of East- their distributive impact is likely to be neutral, as there ern Europe and the former Soviet Union.The EBRD is no evidence that within country inequality has ris- produces similar indicators for the latter two areas. en following their adoption (Dollar 2004), while the Such measures are an extension of an earlier work growth effect is, as mentioned above, positive. How- done by Lora (1997). These synthetic Lora indexes ever, other studies have come to the opposite conclu- suffer from the weaknesses illustrated above, but sion. For instance, an analysis of the impact of over- were compiled according to the same methodology all liberalisation in 18 Latin American countries over and thus allow comparison of the extent of reforms 1980-98 (Berhman et al 2000) found that the reform across countries and over time. In addition, their had a significant short and medium term disequalizing coverage was extended to 1999 by Lora (2001). Each effect, although this effect declined over time. As for policy index is normalized between zero (in the case the impact of specific policy instruments, the study of no reform) and one (in the case of a complete re- found that trade liberalisation did not significantly af- form). The description of each index is as follows: (i) fect inequality, and that specific country outcomes de- trade reform is the average of the mean tariff rate and pended on concrete country conditions. The strongest of tariffs dispersion, (ii) domestic financial reform is disequalizing effect was due to international financial measured as the average value of bank borrowing and liberalisation, followed by domestic financial liberali- lending rates and of reserves to deposits ratio; (iii) sation and tax reform. In turn, a review of the effects international financial liberalisation is the average of liberalisation and globalisation policies during 21 of indexes that reflect the control of foreign invest- reform episodes in 13 Latin American countries, Rus- ments, limits on profit and interest repatriation, con- sia, Turkey, Zimbabwe and India over the last two de- trols on external credits by national borrowers and cades (Taylor 2005) shows that inequality rose in 13 capital outflows; (iv) the tax reform index is equal cases, remained broadly constant in 6, and improved to the average of the maximum marginal tax rate on in two (Chile post-1990 and Costa Rica) i.e. countries corporate incomes and personal incomes, the value where the domestic market and institutional condi- added tax rate and the efficiency of the value added tions were ripe for trade liberalisation. These studies tax; privatization is expressed as one minus the ratio indicate that each policy instrument has a distinct ef- of value-added in state owned enterprises to non-ag- fect on inequality. ricultural GDP; and (v) the general reform index is the simple average of the indexes just described. From a theoretical perspective, some studies emphasize that the standard theory suggests that in countries with 8.3 Results from the literature an abundant supply of labour, trade liberalisation and opening up to FDI and portfolio flows raise exports of Reasons of space do not allow us to carry out a thor- labour intensive manufactures and employment, and so ough review of the literature in this field. Suffice it reduce earnings and income inequality (Wood 1994). here to mention that global cross-country studies have Yet, while this might be true in the standard model, the analyzed the effect of openness on economic growth. stringent conditions under which such theoretical pre- Some of them (Sachs and Warner 1995, Dollar and dictions apply are rarely verified, and different models Kraay 2003 and the literature therein) conclude that are needed to represent the real life situation of coun- trade policy and openness are important determinants tries characterized by widespread market failures (Cor- of economic growth. However, these results have been nia 2004, and the literature therein). Indeed, in many

76 Globalization and Health Knowledge Network Figure 18: Correlation between the real interest rate and the Gini coefficient of income inequality

cases, standard economic theory is unable to predict the distortion of the estimated parameters we expressed growth, inequality and volatility impact of internal and the GDP/c as a GDP index or as a deviation from its external liberalisation, as it is based on simplified models 1960-80 trend, while GDP/c volatility was also ex- pivoting around highly restrictive assumptions which pressed as a deviation from its 1960-80 trend. do not take into account the impact of institutional weaknesses, structural rigidities, incomplete markets, For regressions (1) and (4) in Table 26 on income in- asymmetric information, persistent protectionism and equality, the Reform Index was lagged by five years, as the complexity of trade and finance in a multi-country, reforms normally take time to alter the income distri- multi-goods environment. bution. We also introduced (as done in Behrman et al 2000) the subsequent reform increments over the sub- 8.4 Econometric results sequent five years so as to be able to examine the time profile of the reforms’ impact. In these regressions, the Conscious of the above theoretical and data limita- Reform Index was interacted with the initial value of tions, we tested the impact of the liberalisation and the Gini coefficient (G0), to test if the inequality im- globalisation policies proxied by the Overall Reform pact of the reform is more pronounced in economies Indexes developed by Morley et al. (1999) – integrat- with low initial inequality, or in those that have a high ed with Lora (2001) – for Latin America relative to initial inequality. We have also included a GDP/c in- the period 1980-1999 and by Campos and Horvath dex and its square, to test if the impact of the reforms (2006) – based on Lora’s formulations – for the East- is more or less pronounced in countries with different ern European economies in transition and the former income levels. Soviet Union countries. The test was carried out for the period 1989-2001 on three key determinants of For regressions (2) and (4), which measure the impact health, i.e. GDP/c, income inequality and income of the overall reform index on the GDP/c index or volatility. As the ‘reform indexes’ are not available on its relative deviation from the 1960-80 trend, the for the first part of our sample (1960-80), to avoid a overall reform index was not lagged but was instead

Pathways of transmission and evidence of impact 77 included in quadratic form since – as observed in the of the monetary authorities of the large countries economies in transition - the impact of the reform and the risk premium set by the rating agencies), in- can be concave, i.e. it produced negative effects in the flation (part of which might however be endogenous short-medium term but positive effects over the long to the reforms) and money supply in the Eastern Eu- term. The same approach was followed in regressions rope. In contrast, the external terms of trade appear (3) and (6), which measure the impact of the reforms to be little significant. on the deviation of GDP/c volatility in relation to its 1960-80 trend. Finally, to remove the confounding ef- These results would seem to provide some support to fect of other non policy factors, we included in the re- the hypothesis that – in the two regions considered – gression controls variables such as M2/GDP (money the policy reforms of the last twenty years have had supply), external debt, real interest rate, external terms a negative effect on three key social determinants of of trade, and inflation. For instance, as shown in Fig- health. In the case of inequality the impact appears ure 18, there is a correlation between income inequal- to be somewhat delayed, and to affect most those ity and the real interest rate. countries with low initial inequality and GFP/c. The results also indicate that the control variables includ- Table 26 below presents the results of the econometric ed explain only a small part of the phenomenon at estimates obtained using fixed effect models for the hand. Therefore, with all due caution, one can con- two regions and the periods indicated above. In all clude that there is some initial evidence of the po- its different specifications, the Overall Reform Index tentially pernicious effect of unfettered globalisation appears to generate significant adverse effects on the policies on health. three impact variables selected. In both Eastern Eu- rope and Latin America the Reform Index t-5 has a However, before accepting these conclusions it is significant positive effect on income inequality, as do necessary to consider the arguments put forward in the reform increments over the subsequent five years, a recent paper by Rodrik (2005), who argues that although the more recent policy changes appear to af- standard growth-policy regressions can be often be fect inequality less markedly. In Latin America, the misleading due to a variety of econometric and speci- interaction term between the Overall Reform Index fication problems, including parameters heterogene- and initial inequality is negative, indicating that the ity, omitted variables, model uncertainty, measure- policy reforms will increase inequality in countries ment error and endogeneity. In particular, Rodrik with comparatively low initial inequality. Finally, the notes (p.11) parameters of GDP index and of its square suggest that the impact of the reforms appears to be more im- “… as long as policy interventions are portant in low income countries, as observed for in- not random and their presence responds stance in the case of the European economies in tran- to unobservables, regressing an economic sition where the richer Central European economies performance variable on policy is uninformative suffered smaller rises in inequality. about the degree to which market failures exist, the extent to which policy interventions are Regressions (2) and (4) confirm that economic targeted on them, the effectiveness with which growth is affected by policy reforms in a concave way government policies are deployed, or the extent in Eastern Europe and in an accelerating concave to which policy interventions are used to create way in Latin America. Similar conclusions apply to and divert rents for political purposes.” GDP/c volatility. In this case the negative effect appears to be particularly strong in the initial phase of For these reasons, we are unable to say whether the the reforms. Finally, it must be noted that some of the negative effect of the globalisation reforms are due control variables introduced do appear to be respon- to their premature implementation, or to govern- sible for part of the increase in inequality, slowdown ment rent-seeking behaviour, or to inadequate policy in growth and surge in volatility. This is particularly implementation (though this problem should have the case for the real interest rate (an exogenous vari- been taken care of by the reform indexes developed able reflecting – in an open economy - the decisions by Lora). Yet, while we can only speculate about

78 Globalization and Health Knowledge Network Table 26: Results of the regression of the Lora’s Overall Reform Index on Income inequality, GDP/c and volatility of GDP/c

European economies in Latin America (1980-1999) transition (1989-2001) Income GDP/c GDP volatility Income GDP/c GDP Inequality index (deviation from Inequality (deviation volatility (1) (1989=100) 60-80 trend)a (4) from 60-80 (deviation (2) (3) trend)a from 60-80 (5) trend)a (6) Constant 18.70*** 1.09*** 12.61*** 37.37*** -.5285*** 8.43*** Reform Index ……….. -1.71*** 31.29*** ……….. -.2328*** 8.62** Reform Index 2 ……….. 1.74*** -36.68*** ……….. -.2090*** -10.57*** Reform Index* Gini0 ……….. ……….. ……….. -.1851*** ……….. ……….. Reform Index t-5 15.19*** ……….. ……….. 13.53*** ……….. ……….. Reform increment t 4-5 12.58*** ……….. ……….. 14.27*** ……….. ……….. Reform increment t 3-4 12.42*** ……….. ……….. 11.28*** ……….. ……….. Reform increment t 2-3 10.64*** ……….. ……….. 14.73*** ……….. ……….. Reform increment t 1-2 9.47*** ……….. ……….. 13.58*** ……….. ……….. Reform increment t 0-1 6.40*** ……….. ……….. 12.01*** ……….. ………..

Money supply (M2/GDP) -.0672*** .0019*** ……….. .0179ws .0004 ……….. Total external debt ……….. 1.66*- ……….. 2.36*** -1.08*** ……….. Real interest rate -.0020*** -.00004 -.0273*- ……….. -.0003 .0152** External terms of trade No data No data No data -.0047 -.0003*- .0122ws Inflation -.0025**ws ……….. .0100*** .00008 .000001*** .0006** GDP index 29.08*** ……….. ……….. 14.99*** .6133*** ………… GDP index 2 -14.79*** ……….. ……….. -4.01** ……….. ………….

F statistic 22.41*** 29.81*** 18.60*** 10.08*** 80.12*** 2.39** R square .018 .280 .047 .005 .680 .009 Number of observations 120 127 306 183 191 191 Number of countries 17 17 24 12 17 17

Source: authors’ elaboration based on GHND; Notes: a - the variables have been computed as the difference between the observed values of the variable and those obtained by a prolongation of the 1960-80 trend, divided by the value extrapolated on the basis of the 1960-80 trend. Notes: *** significant at the 1% level; ** between 1 and 5% level; * between 5 and 10 % level; *- significant at 10-15%. ws = wrong sign (i.e. a sign different from that expected on the basis of the theories discussed earlier).

what lies behind the negative association found ses and country studies (as in Cornia 2004) of the between liberalisation-globalisation policies, type suggested by Rodrik himself, in which the theo- poor economic performance and unsatisfactory retical predictions of key economic models (for ex- health trends, it is also evident that – whatever ample the Heckscher-Ohlin model and its corollary) the causes of this impact – the association seems are juxtaposed with the observed effects in a different to be quite robust. Finally, our confidence in the variety of economies where the model conditions are above results derives also from theoretical analy- not verified.

Pathways of transmission and evidence of impact 79 Globalisation and Health: Pathways of Transmission and Evidence of Impact

9. In lieu of conclusions

his paper has argued that: GDP ratios. As a result, national and international financing problems have limited the potentially posi- T tive impact of the transfer of medical discoveries, of (i) an analysis of the health impact of globalisation re- nationally and internationally supported health pro- quires the specification of the pathways (or models) grams such as immunisation and – in some regions through which this impact is felt. Such pathways dif- - of efforts at reducing female illiteracy. The paper has fer substantially from region to region and this re- also argued that technical progress in health appears to quires – at the global level – the development of a have compensated in part for some of these negative multi-pathways model that can take into account the effects and that, for these reasons, policies - concern different effects; ing the production and global dissemination of health knowledge are central to current and future health im- (ii) the determinants of health – particularly those which provements. As for health behaviours, the trends have depend on policy decisions - have often deteriorated been mixed with greater health awareness spreading in in the 1980s and 1990s. Slower growth, higher insta- some regions (such as the OECD) and a passive policy bility, informality and inequality, and sudden changes stand and increased smoking, drinking and spread of have been observed in many countries and years, while nutritionally inadequate Western foods in many liber- access to health services became more dependent than alizing economies where TNCs have now relocated. in the past on out-of-pocket expenditures. These trends are to a large extent related to liberalisation and (iii) because of these trends in their determinants, as well as globalisation policies which have led to systemic in- the impact of AIDS and other shocks such as wars and stability and inequality, and to a weakening of the role disasters (which do not, however, appear as yet to be of the state because of stagnant or declining tax rev- significant at the global level), LEB, IMR and U5MR enue, even in countries with an already low revenue/ improved in the 1980-2000 period at a slower pace

80 Globalization and Health Knowledge Network compared to 1960-80. This occurred despite the many While highly preliminary, these empirical results have favourable ‘dividends’ in the field of demography, tech- to be interpreted, however, in the light of other theo- nology, peace and spread of democracy and market rule retical considerations, and of specific cases studies, as enjoyed by many countries during this period. done throughout this paper. For this reason, we believe there might be some merit in the position that sees the (iv) the econometric estimation of a global ‘multi-pathway unsatisfactory health trends identified above to be re- model’ produced encouraging results. This model is lated also to a premature and acritical application of able to reproduce plausible and significant param- liberalisation-globalisation policies which ignores the eters for many variables – and can therefore be used specific conditions and institutions of each country. to simulate the trajectory of LEB (as well as IMR and U5MR) in 2000, assuming that all their determinants What can then be done? It is quite possible that if behaved during the liberalisation and globalisation era properly ‘managed’, many globalisation policies could as they did over 1960-80. The results of this simu- lead to health gains. Where markets are complete, lation – which like all simulations relies on several competitive and non-exclusionary, regulatory institu- hypotheses – shows that worldwide LEB fell by 1.52 tions strong, asset and income concentration accept- years because of globalisation policies in relation to able, and access to public health services widespread, the counterfactual scenario – although it only stag- many globalisation policies are likely to produce posi- nated if the positive impact of technical progress in tive health effects. But such conditions are met at the health (and the negative effect of AIDS) are taken moment only in a handful of countries. In most of the into account. Clearly the issue of technical progress others weak domestic structures are common, while in health and of the distribution of its benefits has major asymmetries in global market relations and un- become – judging also from our empirical analysis – equal conditions of market access remain dominant quite central to the future of health equity. – as in the case of protectionism in OECD countries, contagion due to financial crises, an unequal distribu- The simulation also shows that globalisation has gen- tion of foreign direct investments and an endless list erated vastly different effects across regions. If we -ex of new conditions (on governance, patents legislation, clude the impact of medical progress, the losers in- social clauses and so on). clude not only sub-Saharan Africa and the transition economies, but also Latin America, South Asia, and, What should the developing countries do in the fu- surprisingly, China and to a lesser extent India where ture? This is not a normative paper but a few broad di- the benefits of very fast growth were more than off- rections can be tentatively put forward. No doubt, the set by mounting instability, inequality and exclusion countries which have been excluded from the benefits from basic health and other services. Among the win- of global markets have a genuine interest in strength- ners one finds the OECD, MENA and, to a lesser ex- ening their human resource base, infrastructure and tent, East Asia. If medical progress over 1980-2000 is macroeconomic balance. These measures – per se – taken into account, then only SSA, some of the transi- would generate health returns and accelerate domes- tion economies and, less markedly, China still record a tic growth, while they would also reduce the barriers lower LEB than in the counterfactual scenario. to their inclusion in the global market. It is equally clear that, for many countries, trade liberalisation and (v) preliminary econometric results of the impact of poli- technology transfers could - in principle - increase ef- cy reforms (as measured by an Overall Reform Index) ficiency, welfare and health. Yet, it is doubtful whether on three key determinants of health (GDP/c, inequal- – under the present increasingly restrictive rules of ac- ity and volatility) in Latin America and the European cess to international markets – any acritical liberalisa- economies in transition would suggest that the reforms tion and globalisation would help them to improve did produce a negative impact, though the specific their market position, economic efficiency, equity and reason for these results (premature liberalisation in health status. For these countries, premature, rapid the presence of widespread market failure, insufficient and unconditional globalisation might generate in policy implementation or rent-seeking behaviour by the short term efficiency, social and health costs that governments) cannot be discerned from our analysis. would worsen economic performance and income

Pathways of transmission and evidence of impact 81 distribution, and erode the necessary political support for a controlled integration into the world economy. Particularly for these countries, a gradual and selective integration into the world economy, linked to the removal of major asymmetries on the global markets and to the creation of new democratic institutions of global governance, is highly preferable to a destabiliz- ing big-bang globalisation. This gradual integration has, however, to be accompanied by steady efforts at the domestic level in the field of a truly universalistic and low cost access to all to basic health, education, and income subsidies for the poor, as well as by stron- ger redistributive policies.

82 Globalization and Health Knowledge Network References

Ahmad, O.B., A.D. Lopez and M. Inoue “The decline in child mortality: a reappraisal”, Bullettin of the World Health Organisation, 2000, 78(10), 1175-1191. Aghion P., E. Caroli and C. Garcia Peñalosa (1999), “Inequality and Economic Growth: The Perspective of the New Growth Theories”, Journal of Economic Literature 37, 1615-1660. Alesina, A. and D. Rodrik (1994), “Distributive Politics and Economic Growth”, Quarterly Journal of Economics. Bach, S. (2006), “International Mobility of Health Professional”, UNU/WIDER Research Papers, 2006/82, UNU/ WIDER, Helsinki. Baltagi, B.H. (2005) “Econometric Analysis of Panel Data”, John Wiley and Sons, Chichester. Beck, T., Demirgüç-Kunt, A., Levine, R. 2000. “A new database on Financial Development and Structure”. World Bank Economic Review 14, 597-605. Behrman, Jere, Nancy Birdsall and Miguel Székely (2000) “Economic Reform, and Wage Differentials in Latin America”, Working Paper of the Research Department n. 435, Inter-American Development Bank, Washington, D.C. Berkman, L. and S. Syme, ‘Social networks, host resistance and mortality: a nine year follow-up study of Alameda County residents’, American Journal of Epidemiology, 1979. 109: p. 186-204. Berry, A. and J. Serieux (2006), ‘Riding the Elephants: The Evolution of World Economic Growth and Income Distribution at the End of the 20th Century’, http://www.networkideas.org/feathm/sep2006/pdf/Berry_Serieux. pdf Bhaduri, A. 2005. “Macroeconomic Policies for Higher Employment in the Era of Globalisation”. Employment Strategy Papers No. 2005/11, Employment Analysis Unit, Employment Strategy Department, International Labour Organization, Geneva. Bicego, G. T., and Boerma, J. T. (1993), “Maternal Education and Child Survival: a Comparative Study of Survey Data from 17 Countries”, in Social Science & Medicine, vol. 36(9):1207-1227 Biggeri, M. (2004), “Has Economic and Social Instability Increased Over the Last Two Decades? Initial Evidence and Interpretation”, mimeo, Department of Economics, University of Florence. Birdsall, N. (2000) ‘Why Inequality Matters: The Developing and Transitional Economies’ (mimeo, February), Carnegie Endowment for International Peace, Washington DC. Black, D., Morris, J., Smith, C. and Townsend, P. (1980), “Inequality in Health: Report of a Working Group (The Black Report)”, Department of Health and Social Security, London Bobak, M., and Marmot M. (1996) “East-west mortality divide and its potential explanations: proposed research agenda” British Medical Journal 312: 421-423 Bourguignon, F. (1998) ‘Crime as a Social Cost of Poverty and Inequality: A Review Focusing on Developing Countries’ (mimeo) DELTA: Paris. Caldwell, J. C. (1979), “Education as a Factor in Mortality Decline: an Examination of Nigerian Data”, Population Studies, vol. 33(3):395-413 Campos, N.F. and Horvath, R. (2006), “Reform Redux: Measurement, Determinants and Reversals”, IZA Discussion Paper No. 2093. Caprio, G and D. Klingebiel (2003) “Database of systemic and borderline financial crises”, The World Bank, Washington D.C.

Pathways of transmission and evidence of impact 83 Chu, Ke-young, Hamid Davoodi and Sanjeev Gupta (2000), “Income Distribution and Tax and Government Social Spending Policies in Developing Countries”, UNU/WIDER Working Papers, No. 214, UNU/WIDER, Helsinki. Cooper, D., McDonald W.D. and Theodossiou I. (2006), ‘The health hazards of unemployment and poor education: The socioeconomic determinants of health duration in the European Union’, Economics and Human Biology, Volume 4, Issue 3, December 2006, p.273-297. Cornia, G.A. (2006), “Pro-Poor Macroeconomics: Potential and Limitations”, Palgrave, London. Cornia, G.A. (2005), Inequality, Growth and Poverty in an Era of Liberalisation and Globalisation, Oxford University Press, Oxford. Cornia, G.A. (2004), “Sudden change and mortality crises”, Popolazione e Storia, 3/2004 Cornia, G.A. (1996), “Labour Market Shocks, Psychosocial Stress and the Transition’s Mortality Crisis”, UNU/ WIDER Research in Progress 4:1996, Helsinki. Cornia, G.A. and S. Rosignoli (2007), ‘Trends in within country income inequality: a reassessment on the basis of the WIID mark2 dataset’, mimeo, University of Florence. Cornia, G.A. and Menchini L. (2006) “ Health Improvements and Health Inequality During the last 40 Years”, UNU/WIDER Research Papers 2006/10, WIDER, Helsinki. Cornia, G.A., T.Addison with S.Kiiski (2005), ‘Income Distribution Changes and Their Impact in the Post-World War II Period’, in G.A. Cornia (2005), op.cit. Cornia, G. A. and Zagonari, F. (2002), “An Econometric Investigation of Changes in IMR and U5MR in AIDS Affected Countries During the Last Twenty Years”, mimeo, UNICEF-IRC, Florence Cornia, G.A. and Paniccià R. (2000), “The Mortality Crisis of Transitional Economies”, Oxford University Press Cornia, G.A. and Danziger, S. (1997), Child Poverty and Deprivation in the Industrialized Countries 1945-95, Oxford University Press, Oxford. Cutler, D., Deaton, A. and Lleras-Muney A. (2005) ‘ The Determinants of Mortality’ Mimeographed, Department of Economics, Harvard University and Woodrow Wilson School and Department of Economics, Princetown University. Deacon, B, M. Hulse and P. Stubbs (1997), “Global Social Policy: International Organisations and the Future of Welfare”, Sage, London. Deaton, A. (2004) Health in an age of globalisation, mimeo, Princeton University, Princeton http://www.princeton. edu/~rpds/downloads/deaton_healthglobalage.pdf Deaton, A. (2001), ‘Health, Inequality, and economic development’, Centre for Health and Wellbeing, Princeton University. Deaton, A. and Drèze, J.(2002), “Poverty and Inequality in India: A Reexamination”, Working Paper No. 107, Centre for Development Economics, Dheli School of Economics, www.cdedse.org/pdf/work107.pdf De Vogli, R. (2004), Change, Psychosocial Stress and Health in an Era of Globalisation, paper prepared for the project ‘Health and Social Upheaval’ supported by the John and Catherine Mac Arthur Foundation, UCLA. Dollar, D. (2001), “Is globalisation good for your health?”, Bulletin of the World Health Organisation, Vol.79 (9): 827-833 Dollar, D. (2004), ‘Globalisation, Poverty and Inequality since 1980”, WPS3333, The World Bank, Washington D.C. Dollar, David and Aart Kraay (2003): Institutions, Trade, and Growth: Revisiting the Evidence. World Bank policy Research Working Paper Ezzati, M, S. Vander Hoorn, C.M.M. Lawes, R. Leach, W.P.T. James, A.D. Lopez, A. Rodgers, C. J.L. Murray (2005), ‘Rethinking the “Diseases of Affluence” Paradigm: Global Patterns of Nutritional Risks in Relation to Economic Development’, PloS Medicine, May 2005, Volume 2, Issue 5, pp.1-9.

84 Globalization and Health Knowledge Network Fox, J. W. (1998),”Gaining Ground. World Well-Being 1950-95”, USAID Evaluation Special Study Report No. 79, USAID, Washington DC Garfield R. and C. Santana (1997), ‘The Impact of the Economic Crisis and the US Embargo on Health in Cuba’, American Journal of Public Health, January Gaziano, T. R. (2007), ‘Reducing The Growing Burden Of Cardiovascular Disease In The Developing World’, Health Affairs, no.1, 2007, 13-24. Godoy, R.A, Reyes-Garcia, V., McDade, T., Huanca, T., Leonard, W., Tanner, S., and Vadez V. (2006), ‘Does village inequality in modern income harm the psyche? Anger, fear, sadness, and alcohol consumption in a pre-industrial society’, Social Science and Medicine, 63 (2006) 359-372. Goesling, B. and Firebaugh, G. (2004), “The trend in International Health Inequality”, Population and Development Review, 30(1):131-146 Grigoriu, C. (2006), “Survival Convergence: Specification Matters”, paper presented at the UNU-WIDER Conference on Advancing Health Equity, 29-30 September 2006, Helsinki. Guillaumont, P., Korachais, C. and Subervie J. (2006), “How Macroeconomic Instability Lowers Child Survival”, paper presented at the UNU-WIDER Conference on Advancing Health Equity, 29-30 September 2006, Helsinki. Gwatkin, D. R. (2000), “Health Inequalities and the Health of the Poor: What Do We Know? What Can We Do?”, in Bulletin of the World Health Organization, vol. 78 (10):3-18 Hawkes, C. (2005), ‘The role of foreign direct investments in the nutrition transition’, Public Health Nutrition, 8(4), pp. 357-365 Hertzman, C. (1995), “Environment and health in Eastern and Central Europe”, A Report for the Environmental Action Program for Central and Eastern Europe. The World Bank. Washington, D.C. House, J., K. Landis, and D. Umberdson (1988), ‘Social relationships and health’. Science, 241: p. 540-545. Jantti M. and S. Sandstrom (2005), “Trends in Income inequality: a critical examination of the evidence in WIID2”, paper presented at the UNU-WIDER Conference on The Future of Development Econpomics, 17-18 June 2005, Helsinki. Jha, R. (2006) ‘Pro-Poor Fiscal Policy in the Globalized Economy’, in G.A. Cornia, ‘Pro-Poor Macroeconomics: Potential and Limitations’, Palgrave, London. Kawachi, I., B.P. Kennedy and R. Wilkinson (1999), “Inequality and Health: a Reader”, The New York Press, New York. Kaminsky, G.L. and Reinhart, C. (1999), “The Twin Crises: The Causes of Banking and Balance of Payments Problems”, The American Economic Review, vol. 89, n. 3, pp. 473-500 Kornai, Janos (1992) The Socialist System: The Political Economy of Communism. Princeton University Press, Princeton N.J. Labonté, R and T. Schrecker (2006), “Globalisation and social determinants of health: Analytic and strategic review paper” (revised March 11, 2006), mimeo. University of Ottawa. Lamar, J (2000), ‘Suicides in Japan reach a record high’, British Medical Journal, 321: 528 Lane, P., Milesi-Ferretti, G.M. 2003, “International Financial Integration,” IMF Working Paper No. 03/86 Livi-Bacci, M. (2000), “Mortality Crises in Historical Perspective: The European Experience”, in Cornia, G.A. and Paniccià R. (2000), op. cit. Lora, E. 2001. “El crecimiento económico en América Latina después de una década de reformas estructurales”. Washington, DC, Estados Unidos: Banco Interamericano de Desarrollo, Departamento de Investigación. Documento mimeografiado.

Pathways of transmission and evidence of impact 85 Lora, E. 1997. “Una década de reformas estructurales en América Latina: qué se ha reformado y cómo medirlo”. Working Paper 348. Washington, DC, Estados Unidos: Banco Interamericano de Desarrollo, Departamento de Investigación. Lu Aiguo (2006), ‘Transition, Inequality, Stress and Health Status in China’, paper prepared for the project ‘Heath amnd Social Upheavals’ sponsored by the John D and Catherine T MacArthur Foundation, Institute of World Economics and Politics, Chinese Academy of Social Sciences, Beijing. Marmot, M and M. Bobak (2000) ‘ Psychological and biological mechanisms behind the recent mortality crisis in Central and Eastern Europe’ in Cornia and Paniccià (2000), op. cit. McKeown, T. (1976), The Modern Rise of Population. London: Edward Arnold. McMichael, A., McKee, M., Skolnikov, V. and Valkonen, T. (2004), “Mortality trends and setbacks: global convergence or divergence?’, The Lancet, Vol. 363(9415): 1155-1159 Meyer, D. (2001), ‘Convergence Clubs in Cross-Country Life Expectancy Dynamics’, WIDER Discussion Paper N. 2001/134, UNU WIDER, Helsinki Micklewright, J. and Stewart, K. (1999), ‘Is the Well-being of Children Converging in the European Union?’, The Economic Journal, 109(459): F692-714 Milanovic, B. (1998), Income, Inequality, and Poverty during the Transition from Planned to Market Economy, The World Bank, Washington D.C. Minujin, A. and Delamonica, E. (2003), “Mind the Gap! Widening Child Mortality Disparities”, Journal of Human Development, Vol.4 (3) Morley, S., R. Machado and S. Pettinato. 1999. “Indexes of Structural Reform in Latin America.” Serie Reformas Económicas. Santiago, Chile: Economic Commission for Latin America. Pasinetti, L. (1974), Growth and Income Distribution: Essays in economic theory, 1974. Cambridge University Press, London. Prasad, E., K.Rogoff, Shang-Jin Wei and M.Kose (2003) “Effects of Financial Globalisation on Developing Countries: Some Empirical Evidence”, IMF Occasional Paper 220, IMF, Wadhington D.C: Preston, S. H. (1976), Mortality Patterns in National Populations, Academic Press, New York Ramey G. and Ramey V. (1995) “Cross-Country Evidence on the Link between Volatility and Growth” American Economic Review. 85 5. Reddy, S.and J. Vandemoortele (1996), User Financing of Basic Social Services: A Review of Theoretical Arguments and Empirical Evidence, UNICEF, EPP Working Paper, New York, mimeo. Rodrik, D. (2005). “Why we learn nothing from regressions economic growth on policies”. Harvard University. Rodrik, D. (2003), “Growth Strategies” Working Paper No. 10050, National Bureau of Economic Research, Cambridge, MA. Rodrik D. (1999) “Where Did All the Growth Go ? External Shocks, Social Conflict and Growth Collapses”, Journal of Economic Growth 4, 4. Rodrik, D., Subramanian, A, and Trebbi, F. (2002). “Institutions Rule: The Primacy of Institutions Over Geography and Integration in Economic Development”. Revised version (October 2002) of the paper on Rodrik’s home page. Ruzicka, L., Wunsch, G. and Kane, P. (1989), Differential Mortality: Methodological Issues and Biosocial Factors, Clarendon Press, Oxford Sachs, J.D., and A. Warner. 1995. “Economic Reform and the Process of Global Integration.” Brookings Papers on Economic Activity 1995(1): 1-95. Sen, A. and Himansu (2004), “Poverty and Inequality in India, 1 and 2”, Economic and Political Weekly, September 18 and 25, 2004.

86 Globalization and Health Knowledge Network Sen, A. K.(1997), “Inequality, unemployment and contemporary Europe”. International Labour Review 136 (2) Summer. Shapiro, Judith (1994). The Russian Mortality Crisis and its Causes. Working Paper No 99. Stockholm: Östekono- miska Institutet. Shkolnikov, V. M.; Cornia, G.A. (2000), “Population crisis and rising mortality in transitional Russia”, in Cornia, G.A. and Paniccià R. (2000), “The Mortality Crisis in Transitional Economies”, Oxford University Press Smeeding, T. (2002), “Globalisation, Inequality and the Rich Countries of the G-20: Evidence from the Luxemburg Income Study (LIS)” Paper prepared for the G-20 Meeting on Globalisation, Living Standards and Inequality: Recent Progress and Continuing Challenges, Sydney Australia, May 26-28, 2002 Stiglitz, J. (1998) “Broader Goals and More Instruments: Towards a Post-Washington Consensus”, UNU/WIDER Annual Lecture, n.2, UNU/WIDER, Helsinki. Székely, M. and Hilgert, M. (1999), “The 1990s in Latin America: Another Decade of Persistent Inequality”, Working Paper 410, Research Department of the Inter-American Development Bank, Washington, DC. Taylor, L. (2005), “External Liberalization, Economic Performance, and Distribution in Latin America and Elsewhere”, in Cornia G.A. (2005) op.cit. Tchernina, N. (2000), Rising unemployment and coping strategies: The case of the Novosibirsk oblast in Russia”, in Cornia and Paniccià (eds.) 2000, op.cit. Tiberti, L. (2006), “Changes in Health Financing and Health Status: The Case of Viet Nam”, paper presented at the UNU-WIDER Conference on Advancing Health Equity, 29-30 September 2006, Helsinki. UNAIDS (2006), Report on the global AIDS epidemic, UNAIDS: Geneva UNCTAD (2003), “Commodity issues, including the volatility in commodity prices and declining terms of trade and the impact these have on the development efforts of commodity-dependent developing countries” background notes, trade and development board, Fiftieth session, Geneva, 6-17 October 2002. United Nations Population Division (2002), World Population Prospects: the 2000 Revision, United Nations, New York. Van der Hoeven R. and C. Saget (2004), Labour Market Institutions and Income Inequality:What are the New Insights after the Washington Consensus?” in Cornia G.A. (2005) op.cit. Van Doorslaer et al. (2006), “Effect of payment for health care on poverty estimates in 11 countries in Asia: ana analysis of household survey data”, The Lancet, Vol 368, October 14, 2006, pp.1357-1364. Venieris Y. P., D.K. Gupta (1986), “Income Distribution and Sociopolitical Instabilitty as Determinants of Savings: A Cross-sectional Model”, Journal of Political Economy vol 94 (4), 873-883. Wagstaff, A. (2000), “Socioeconomic Inequalities in Child Mortality: Comparison across Nine Developing Countries”, in Bulletin of the World Health Organization, vol.78(1):19-29. Wagstaff, A. and Cleason, M. (2004), “Raising to the Challenge”, the World Bank, Washington D.C. Wilkinson, R. G. (1996), ‘Unhealthy Societies: The Afflictions of Inequality’, Routledge, London & New York Wilkinson, R..G. (2000), ‘Mind the gap: hierarchies, health and human evolution’ Widenfeld and Nicolson, London. Wilkinson, R. G. and Pickett (2006), ‘Income inequality and population health: A review and explanation of the evidence’, Social Science and Medicine, 62, pp. 1768-1784. Wilson, C. (2001), “On the scale of of global demographic convergence”, Population and Development Review, 27 (1):155-171. Wood, Adrian (1994), North-South Trade, Employment and Inequality, Clarendon Press, Oxford.

Pathways of transmission and evidence of impact 87 Wooldridge, J.M: (2001) “ Econometric Analysis of Cross Section and Panel Data” , The MIT Press, Cambridge, Massachussets. World Bank (2006) World Development Indicators 2006 Cd-Rom, World Bank, Washington. World Bank (2004) World Development Indicators 2004 Cd-Rom, World Bank, Washington. World Bank (2000) World Development Report 2000-2001. Attacking Poverty, Oxford University Press, New York. World Bank (1999), “Curbing the Epidemic: Governments and the Economics of Tobacco Control”,World Bank. Washington, D.C. World Health Organisation (2006), “Working Together for Health: the World Health Report 2006”, WHO, Geneva. Zhang, X. and Kanbur,R. (2003), ‘Spatial Inequality in Education and Health in China’, mimeo, IFPRI, Washington and Cornell University.

88 Globalization and Health Knowledge Network Publications of the Globalization Knowledge Network

Towards Health-Equitable Globalisation: Rights, Regulation and Redistribution; Report to the Commission on Social Determinants of Health Ronald Labonté, Chantal Blouin, Mickey Chopra, Kelley Lee, Corinne Packer, Michael Rowson, Ted Schrecker, David Woodward and other contributors to the Globalization Knowledge Network.

Globalisation and health: pathways of transmission, and evidence of impact Giovanni Andrea Cornia, Stefano Rosignoli, Luca Tiberti

Trends in global political economy and geopolitics post-1980 Patrick Bond

Globalisation and policy space Meri Koivusalo, Ted Schrecker

Globalisation, labour markets and social determinants of health Ted Schrecker, with assistance from Daniel Poon

Trade liberalization Chantal Blouin, Sophia Murphy, Aniket Bhushan, Ben Warren

Aid and health Sebastian Taylor

Globalisation, debt and poverty reduction strategies Michael Rowson

Globalisation, food and nutrition transitions Corinna Hawkes, Mickey Chopra, Sharon Friel, Tim Lang, Anne Marie Thow

Globalisation and health systems change John Lister

Globalisation and health worker migration Corinne Packer, Ronald Labonté, Denise Spitzer

Globalisation, water and health J. Zoë Wilson, Patrick Bond

Intellectual property rights and inequalities in health outcomes Carlos Correa

Globalisation, global governance and the social determinants of health: A review of the linkages and agenda for action Kelley Lee, Meri Koivusalo, Eeva Ollila, Ronald Labonté, Ted Schrecker, Claudio Schuftan, David Woodward

Pathways of transmission and evidence of impact 89